The role of the committee is to explore artificial intelligence and the opportunities and challenges it presents. We have been looking at the relationship between AI and the State and how State agencies will deploy artificial intelligence. Today, we are moving on to the topic of justice and law enforcement.

I am sure our witnesses are well aware of the terms and conditions of appearing before Oireachtas committees, so I will forgo the long preamble. We are pleased to be joined Assistant Commissioner Ms Angela Willis, who has responsibility for organised and serious crime, and Mr. Fintan Brady, director of ICT, from An Garda Síochána, and by Dr. Elaine Byrne, chairperson, Ms. Helen Hall, chief executive, and Mr. Cian Finn, senior manager, from the Policing and Community Safety Authority. I welcome the witnesses. We will take their opening statements and then we will then go to questions and answers from the members. I invite Ms Willis to make her opening statement.

I thank the committee for the invitation to address it. I am the assistant commissioner with responsibility for organised and serious crime in the Garda Síochána. I am accompanied by my colleague Mr. Fintan Brady, who is our director of ICT.

As we implement our three-year strategic plan, our mission remains to keep people safe. However, the environment in which we pursue that mission is being fundamentally altered by rapidly changing developments in new technologies, including artificial intelligence. I will outline how we are balancing the requirement to keep people safe in that changing environment, including the urgent need for An Garda Síochána to be appropriately supported with AI-enabled technology in the context of our absolute commitment to human rights, transparency and the human-in-the-loop principle.

Artificial intelligence is changing how policing functions, how crime is committed, how evidence is generated, how victims are targeted and how disinformation and harmful content can be created and circulated. It is also changing the tools available to law enforcement. An Garda Síochána is transitioning from a traditional-style police service to a data-led, digitally enabled organisation. We are no longer just policing physical streets; we are policing a digital landscape where crime moves at the speed allowed by fibre optic cable. In our three-year strategic plan we have set out that in facing the strategic challenges of the future we will continuously work to understand the threats and opportunities presented by artificial intelligence, disruptive technologies and algorithmic models. With the commencement of the relevant legislation, we will seek to create operational efficiency and increased public safety through automating the onerous task of reviewing hundreds of hours of CCTV footage or thousands of images of online child sexual exploitation material. Leveraging available technology will also support the health and well-being of our members, ensuring that their exposure to harmful online content is minimised as much as it can be through using AI tools to process and filter large volumes of online abuse material.

Looking ahead, our approach is structured around three pillars: governance first and innovative capability development with strong safeguards and transparency. We will implement mandatory data protection and human rights impact assessments for any proposed use of AI which will also be subject to the development of a code of practice in line with the requirements of the legislation. While AI can provide assistance to policing, decisions on how the information is utilised will always have human input. We are guided by both the EU AI Act and the national guidelines for the responsible use of AI in the public sector which emphasise that AI must serve the public good. We recognise that in order for AI to be effective in policing, it must first be trusted. We will integrate the recommendations made by this committee to ensure that every algorithm is governed by law, every search can be audited and every human in the loop remains accountable.

While artificial intelligence is often considered a new or relatively new technology, it should be noted that AI in policing is not a new concept. AI technology is embedded in many police services both within and outside the EU. From our international engagement in the investigation of organised and serious crime, we see the benefit of using AI systems to analyse large data sets, cluster relevant information and detect relationships to prioritise evidential opportunities. We strongly support the development of a co-ordinated all-of-government approach to AI. Clear standards, shared infrastructure and consistent oversight will benefit the public and front-line services alike, ensuring the public continue to have trust.

I emphasise that technology will not solve the core challenges of policing. Trust, professionalism, community engagement and accountability remain the foundation of effective policing. Our plans for AI are rooted in these principles. We are committed to working with this committee, regulators and the public to ensure that any use of AI in policing in Ireland reinforces public safety and strengthens public trust. We must not allow a digital sanctuary for criminals to exist. By adopting these tools responsibly, we ensure that An Garda Síochána remains a modern, effective and above all a trusted protector of the Irish public. I welcome questions and the opportunity to discuss our approach in more detail.

I now invite Dr. Elaine Byrne from the Policing and Community Safety Authority to make her opening statement.

I thank the Chair and members for the invitation to meet with them today. The Policing and Community Safety Authority, PCSA, was established on 2 April 2025 under the Policing, Security and Community Safety Act 2024. We are an independent statutory body responsible for oversight of An Garda Síochána in respect of the delivery of policing services. We comprise an authority of nine members who are supported by an executive team under the CEO. We are the successor organisation to the Policing Authority and the Garda Inspectorate.

Our core objective is to oversee and assess the performance of An Garda Síochána in providing policing services to communities across Ireland.

In respect of the introduction and use of AI and other emerging technologies in policing, we recognise that the overarching legal framework and pivotal decisions on whether and how to utilise them are primarily matters for the Oireachtas through the establishment of underpinning legislation. Thereafter, the manner in which they are introduced and given operational effect is the prerogative of the Garda Commissioner, the Garda board, the Department of Justice, Home Affairs and Migration and wider Government. Our involvement in this process is through the provision of informed advice on policing matters and as a statutory consultee.

As the independent oversight body, our primary role in this space is to monitor the implementation and use of these systems by An Garda Síochána in its provision of policing services. As such, our strategy for 2026 to 2028 contains a priority action to scrutinise the adoption of new and evolving technologies by An Garda Síochána to ensure their use is appropriate and proportionate. In practice we will place a key focus on how the use of AI is consistent with the statutory functions of An Garda Síochána and policing principles, namely, that its use is necessary, proportionate and lawful; that it protects, vindicates and does not inappropriately infringe on human rights; that it supports effective administration of justice; that it is understood by and considers the experiences of the public; and that it is undertaken in an ethical and professional way.

We are also closely following international policing research, developments and best practice to inform our work. Current and future AI adoption is a critical matter for society with all sectors currently grappling with its potential benefits, risks and impacts. Its introduction is largely a case of when rather than if. Policing is no different. Considering the use of AI in the context of policing is different given the power vested in police services. In particular, given that An Garda Síochána is both a national police and security service, there are heightened considerations to be given to the safeguards and governance around its introduction and use. The adoption of AI by An Garda Síochána has already commenced and is likely to be further facilitated through new and planned legislation such as the Garda Síochána (Recording Devices) (Amendment) Bill 2025 and that concerning live facial recognition. We recognise that emerging technologies including AI can have significant benefits for policing and the service offered to the public, including victims of crime. Emerging technology can assist in the prevention, investigation and detection of crime. It can also contribute to the more efficient management of the Garda organisation. It is also important to remember that AI and emerging technology can be and currently are utilised and exploited by those engaged in criminality and those who seek to threaten the security of the State. This can include the spread of misinformation and disinformation, the creation of illegal images and the execution of widespread cyber fraud activity.

In that regard, policing must be adequately prepared and equipped to respond to evolving criminality. In these contexts, we seek to prioritise the roll-out of emerging technology. We need to be clear and realistic about where and when emerging technology such as AI might provide a benefit to policing rather than engage in its adoption for adoption’s sake. We believe that focused consideration needs to be given to the following: key learnings and experiences Ireland can draw from other policing jurisdictions that have been early adopters; the necessary legislative and policy infrastructure that might be required to facilitate its introduction and govern its use; the clarity that exists within An Garda Síochána and is communicated to the public as to when, how and within what parameters the technology will be used; how it can be used to vindicate human rights and provide an enhanced policing service for the public and for victims; and the capacity within An Garda Síochána to adopt the technology. In this context, in setting statutory policing strategies for 2026 to 2028, we have sought that An Garda Síochána publish a clear, accessible and transparent framework that sets out the safeguards it will use when considering introducing and using emerging technology.

This will continue to be a priority area for oversight in the coming years. We welcome the opportunity to discuss that with the committee today.

I thank Dr. Byrne. We move to the members for questions and answers. A speaking rota has been agreed. Each member will have seven minutes. I remind members to be succinct with their questions to allow enough time for answers. The first on the rota is Senator Gareth Scahill. Are you on the campus, Senator? I can see you on the screen.

No, I am not on the campus.

You can join us when you arrive. I call Deputy Geoghegan.

I thank the witnesses for participating in the committee. I might start with Assistant Commissioner Willis. Last week, we had the head of Coimisiún na Meán, Dr. Godfrey, with us. Under questioning, he said that, at the moment, it is not a criminal offence under Irish law to deploy an AI system that can be used to create intimate imagery of real people without their consent, or that is capable of producing child sex abuse material. Is that statement of the law correct?

I thank the Deputy for the question. Child abuse imagery is governed by the Child Trafficking and Pornography Act. The development of child abuse imagery, whether it is of a genuine child or AI-generated material, does constitute a criminal offence.

Dr. Godfrey was trying to draw a distinction between the producer of that imagery and the system that enabled the production of that imagery. Is there a distinction in Irish law between the enabler of the production as distinct from the producer?

Our investigations historically focused predominantly on whoever generated that material. However, in our investigations, we will go wherever the evidence leads us. If somebody knowingly deployed a tool where the purpose was to create those images, whether a prosecution was initiated would ultimately be a matter for the Director of Public Prosecutions, DPP. However, we would certainly believe it would be worth investigating, bringing all of the evidence together, which can be quite challenging, as the Deputy will appreciate. Sometimes, the people who deploy those tools may have to self-incriminate, and that can be challenging for us in trying to get that evidential piece, if you like. It can transcend multiple jurisdictions.

Without prejudice to any ongoing, live criminal investigation, could the systems themselves, and the persons who have responsibility for the control of those systems, be subjected to criminal investigation if and when child sex abuse material, or non-consensual adult image material, has been produced using their systems?

They could certainly be the subject of criminal investigation. As I said, whether a prosecution would follow from that would be dependent on the DPP considering that evidential proof was available. Certainly, in order to initiate an investigation, there is ample scope if somebody is aiding or abetting the generation of such material when it relates to a child.

Dr. Godfrey's point in the context of the world he operates in, which is perhaps more in the civil regulatory space, as Coimisiún na Meán, was that one of the prohibited practices contained in the AI Act does not currently include the deployment of a system that would facilitate the production of non-consensual adult imagery or child sex abuse material. He recommends, as does the AI council, that we amend the AI Act at a European level to include this as a prohibited practice. How relevant, if at all, would that change be in a criminal law context?

We would welcome any additional safeguards that can be put in place, as well as obligations on the operators of the systems that facilitate and allow material to be generated.

However, as we currently stand in relation to child abuse imagery, we certainly have ample opportunity to investigate and take all reasonable steps to secure any evidence that might be useful in a prosecution.

Without prejudice to any ongoing live investigation, does Ms Willis foresee current learnings from investigations? This is a relatively new area in some respects which could result in the Legislature, us in the Oireachtas in an Irish context, being asked to amend or look at the existing legislation that the Garda has at its disposal to carry out investigations.

We will always welcome new legislative abilities. As the Deputy has said, this is a new and emerging space for us. However, in relation to child abuse material, I am confident that the tools currently available to us allow us to investigate such actions. It is slightly different when talking about images that are generated of adults in that, going back to Coco's Law, we need a complainant to actually come to us with a complaint in relation to the material.

Is non-consensual adult imagery a trickier area in terms of us applying Coco's Law in this new world we operate in with AI systems? Has it now become even more complicated in the face of these AI systems and what they can produce?

Yes.

Does Ms Willis anticipate that An Garda Síochána in its engagement with the Minister will explore potential new legislative proposals to strengthen Coco's Law?

Yes. I am aware that is already under consideration.

Dr. Byrne indicated that the adoption of AI by An Garda Síochána has already commenced. Ms Willis's statement referred to the gathering of data in organised crime settings. What does that commencement of AI currently look like in An Garda Síochána?

We do not have any AI systems deployed within An Garda Síochána currently. We see enormous opportunity for the deployment of AI systems to support our members. We were discussing earlier the potential areas we could look at as a starting point. The first thing is we need to develop the strategy around the deployment of AI within the organisation. That will be the first step. We have committed to doing that in our three-year strategic plan. We look at projects that will be low risk initially. We will then engage with international partners but also people here. I know representatives from the Revenue Commissioners appeared before the committee a number of weeks ago. They are considerably ahead of us in terms of their use of AI technology in support of the people working in Revenue, for example around a knowledge hub. Having access to all the relevant legislation, policies and guidelines in one area could provide instant accurate information for our members. We are a complex or organisation and we have many policy areas. I have over 225 policy areas assigned just to me, as one assistant commissioner. Policy is complex. We are a growing organisation with lots of new people coming into it. Something like a knowledge hub, which our members could tap into to inquire about the requirements, guidelines and policies about any particular incident they might be dealing with, would give them a system providing reliable information at their fingertips. With many incidents we deal with, there could be several policy areas impacted in one stop-and-search, for example. It would be useful if there was a knowledge base that a member could tap into to get accurate information to inform their decision-making on what they need to do, but obviously always with a human making the ultimate decision.

I thank the witnesses for their opening statements. I wish to pick up on the point the assistant commissioner mentioned on the potential for members to use AI systems.

Would AI be given access to PULSE? Is that something that would be envisaged at some point? Would that be considered low risk?

No. I am referring to information that is available publicly and some internal documents which do not contain personal data. We would not rule out having tools in the future that could search through PULSE. One such example is economic crime.

I have a lot of questions. The responsible use of AI in the public sector draws from the AI Act. There is little within the sphere of the Garda that will be low risk. It will all be high risk because it is such sensitive data. Who has responsibility in the organisation for the governance of AI?

Our chief corporate officer sits at the top of the ICT side of the house. My colleague, Fintan Brady, is the director of ICT. We do not currently have a CIO. However, one has been recruited and the post will be filled later this month.

In preparation for the deployment of AI in policing practices in Ireland, the Garda is monitoring the current landscape in terms of that governance. I am particularly interested in the digital omnibus packages that are going through at the moment because they seek to draw back some of the protections that were to be introduced by the AI Act. Ms Willis has shared with us that she would welcome additional guardrails. Does she have concerns about the omnibus packages and how they might roll back on those protections?

I do not because in our role, trust is the most important aspect of everything we do. We have a high trust rate. Our last public attitudes survey showed that 92% of the public trust us. That is important for us. In anything we do, we have to ensure we have the support of the public we serve.

Ms Willis does not think that the digital omnibus package threatens that trust.

I do not think it does. It will streamline things and make them simpler. The intention is that it will be a simpler process. We will always carry out data protection and GDPR impact assessments. Regardless of what the technology is, we will take a range of basic steps. Regardless of what the omnibus package says, we will still take those steps. We are subject to comprehensive-----

The Garda goes above and beyond what the AI Act requires.

Yes.

Why, then, does it make sense to deregulate this? Ms Willis has used terms like "simplify" and "streamline". I see it as deregulation. It is quite simple and clear to me that that is the case. There was a regulatory proposal coming through the AI Act and it will now be downgraded with clear protections being lost. Why does the Garda support that if Ms Willis said it will go above and beyond?

I am not saying we support it. We, as An Garda Síochána, implement whatever regulation and legislation the Oireachtas implements.

I will finish because I want to move to the PCSA. It was said that AI is being deployed in other jurisdictions. One of the key areas where we know it is being deployed is in the use of facial recognition. There is AI that underpins facial recognition, which, in other jurisdictions, has proved to be very problematic. Recent UK research identified that there are false positives on Asian and black individuals in its systems. I appreciate the assurances around "the human-in-the-loop principle" and the other phrases that were referenced, such as "safeguards" and guardrails. Do the witnesses have concerns about such technology being deployed? I am concerned that those biases are baked in from the outset and it is very difficult to roll them back. Are we and the technology ready to be deployed in that space?

In 2019 and 2020, Canada had issues around Clearview AI in respect of facial recognition. There was quite a large scandal there. Ireland coming late to the party is no bad thing. We can learn from other jurisdictions, in particular Canada.

Has the technology learned and improved on the back of these scandals?

As regards the purpose of the oversight body in the Canadian context, the principles it has adopted are ones that we would seek to adopt. We are conscious of the mistakes made in that jurisdiction around the introduction of certain technology when it was not communicated to the oversight body beforehand where there was communication to the public. Essentially, with our principles, we would be probing these technologies in a device-neutral way, regardless of whether it is facial-recognition technology or tasers or-----

I just want to clarify this. The oversight Dr. Byrne is envisaging is after deployment. Is that correct?

No, not at all. That was exactly the mistake that happened in the Canadian context.

Okay.

In that context, what we are already doing in our monthly meetings with the build leadership team is seeking information all the time around the different technologies on a principles'-based-----

Does the authority have the power to withhold approval or sanction for deployment of AI in An Garda Síochána?

That is an operational matter for An Garda Síochána. As an oversight body, our job is to ensure that the technology being deployed is necessary, proportionate, human-rights based and aligns with legislative principles.

If the authority feels that is not the case, what does it do? Does Dr. Byrne just express concern in those public meetings?

Absolutely. We would be communicating that, as is our obligation, to the Minister for justice.

Yes, but An Garda Síochána cannot actually be prevented from using the technology, even if the authority does not feel that it is appropriate for use in this jurisdiction. Is that correct?

In my two or three years as chair, it is difficult for me to find an example of An Garda Síochána proceeding with a particular policy that we were against.

I will focus on road safety and our future use of AI in assisting us with it. In my community we have a number of what I will call car enthusiasts who gather to do donuts, rev their engines and all those kinds of things, keeping communities awake. I met local gardaí and they have been very helpful. They told me about the operations they are doing, which can be successful on the night but very difficult to organise, and in a couple of months the car drivers will start gathering again. It is obvious that they are organising these meet-ups on Snapchat or other social media platforms. At the time, the Garda told me that the cybercrime unit is overwhelmed with child sexual abuse material, CSAM, and fraud and does not really have the capacity to monitor social media for the likes of meet-ups, where dangerous driving might take place. I would like to get the assistant commissioner's thoughts on that. Are more resources needed for the cybercrime unit?

Regarding the capacity of the cybercrime bureau, the Deputy will be aware that a number of years ago we had significant backlogs. We deployed four regional hubs to help address that. We have invested significant resources in the Garda National Cyber Crime Bureau over the last number of years. I am happy to say that we have a very small and manageable backlog in the bureau at the moment.

If meet-ups are being arranged on Snapchat, it would require intrusive access into devices belonging to individuals. For us to engage in any attempt to access information such as communication, there is a very high bar set. I can tell the Deputy that a meet-up such as she describes would not meet that threshold, because the bar is really high. If we wanted to do what the Deputy is calling for, it would be intrusive surveillance because we would be monitoring conversations between people if they were arranging a meet-up. I acknowledge the harm that this activity causes to communities and to people's right to have safe and peaceful communities to live in. Even if we had an unlimited amount of resourcing in the cyber crime bureau, it is not something that we could engage in.

I will stick to the same topic. When I posted about this on my social media, I got trolled by youngsters who were having fun in their own minds.

I said to one of them, in particular, that he might think about turning his privacy settings on before he comments on a TD's post, while on his own social media there are videos of him using his phone while driving, doing donuts and changing his number-plates to fake number-plates. When I brought this up, I was told that even where there are screen recordings of somebody's social media or that person is publishing on their own Instagram or Facebook account, the Garda cannot use that as evidence. The Garda has to get that material straight from the social media companies. Will the assistant commissioner explain that?

We can certainly initiate an investigation. I am happy to speak to the Deputy after this meeting and get some specific details. If she has evidence of that kind of behaviour, we would certainly seek to follow up on it and bring it as far as we can from an evidential-----

I am not talking about me specifically. I do not want to go out targeting my constituents and trolling their social media. Obviously, we cannot go onto Snapchat and monitor private chats, but if one of the main motivations for these meet-ups is recording the event and then putting that material on social media, can anything proactive be done as regards searching social media for evidence of dangerous driving and then trying to convict? I am not the fun police but this is also a road safety issue.

We do open source inquiries and whatever we come across, we will follow up on it. Specifically on the road safety issue the Deputy referred to, we have no specific operation targeting that at the minute. It is certainly something that we can consider because I agree that this is a road safety issue above anything else. It is not fun to be out in the community doing donuts in cars that have been modified, etc. I will take this back and talk to my colleagues in the roads policing community engagement bureau to see if we can follow up on it.

Sticking with road safety, Ms Byrne mentioned that she is closely following international policing research developments and best practice to inform her work. Has she seen cases where AI is being used in road safety? In Ireland, it is not yet possible to upload dashcam footage to the Traffic Watch reporting form. I hope that will be in place soon. Has Ms Byrne seen AI being used for road safety in other jurisdictions?

I am sorry to hear that Deputy Keogh has had that experience. Lots of women on social media have a similar experience. I know what it is like. I ask my colleague Ms Hall to comment.

I thank the Deputy for the question. I do not have any particular examples of AI being used. It is, however, something that could be used positively in future. It is like everything we are talking about today. When we are talking about AI, it is such a spectrum. It is everything from minor issues, such as looking at policies, which the assistant commissioner talked about, right up to the very intrusive piece.

The big piece of work for the authority, as I would see it, goes back to what Deputy Gibney talked about around the governance piece. What we would be encouraging An Garda Síochána to do is use its might. It is a very large police service, but it comes right back to the point of procurement. Do you actually need this tech in the first place and, if so, what are the intrusive powers and how do you procure it and demand from the suppliers in advance that the algorithms should be open? It then goes beyond that. If you bring governance to the next issue of AI, it is around making sure the tech is open and the user, An Garda Síochána, has the information about how the algorithms are used. However, there is another piece for the communities in terms of how the Garda is going to deploy this tech because it is in its infancy. The way in which the Garda is going to use the tech needs to be very clear. It needs to be done in a coherent way because it is quite nebulous for us all. We are not experts on this. I see our role as making sure that the public can be easily and simply informed of when the Garda is using AI, and any such technology or any emerging tech, and that it is very clear when it is going to use it and how is going to mind it within the organisation. As we have seen over the years, it is challenging for the Garda to implement that governance. When it gets advice on human rights, for example, that should be very easily told to the public in order that people know their rights are being protected appropriately.

I thank the witnesses for attending. I will start with the assistant commissioner. Another committee was told that there are 200 active investigations into child sexual abuse-related images generated by the X AI chatbot, Grok. Are any of those investigations into Grok or its senior management?

As in, I could envisage investigations into someone who uses Grok to create child sexual abuse imagery or other abuse imagery of women, but are any of these investigations into Grok as the company that is effectively creating the tools for this imagery creation?

I thank the Deputy for the question. We have commenced 244 investigations, and about 50% of those relate to child abuse imagery or child pornography, as is defined in the Child Trafficking and Pornography Act.

As I alluded to earlier in reply to Deputy Geoghegan, we will go wherever the evidence takes us, which might first be a user, but then more broadly, depending on what evidence and co-operation we get from a particular platform and what evidence we can glean from that. I said earlier that can be another jurisdiction so it can be quite a complicated and lengthy process. However, we will take those as far as we can from an evidential perspective. Obviously, we will then bring that information before the DPP. We are not ruling anything out.

Effectively, An Garda Síochána is investigating this, which may lead to prosecutions of Grok or Grok senior management, as opposed to or as well as potentially, individuals who have used those tools. However, they have not settled on whether this is a definite result yet.

No. The thing to say is these are also quite lengthy investigations because of that transnational nature of a server being in one country and generated in a different country. The user who has generated the material can be elsewhere, and that is the nature of the online space we now find ourselves policing.

As I said, we will go wherever and we will spare no effort in trying to identify all the different opportunities to bring perpetrators before the courts.

I do not know if Ms Willis has seen that last week or the week before, it was reported that the UK police used live facial recognition technology in Holyhead. Were gardaí made aware of this by its counterparts in Britain before it happened?

I am personally not aware of that. I would not like to say the organisation was not made aware of this because it might not necessarily come to my door.

Does Ms Willis have any concerns about this? It appears they are using it live. This is meant to be a common travel area, yet we effectively have some form of border control happening in the form of facial recognition technology.

As regards live facial recognition technology, we do not have the legislative basis to do that here at present. We heard earlier about the progression of the An Garda Síochána (Recording Devices) (Amendment) Bill, which is currently going through the process. By and large, that caters for retrospective biometric recognition and only in very minute areas can we use it in a live setting, such as State security or missing persons.

Again, it is very heavily regulated and we do not have it at the minute, so it is not something we would consider using.

Regarding An Garda Síochána plans, obviously Government is pushing this legislation through, and unless it changes its mind, or Government falls, this legislation is likely to pass. How advanced are the organisation's preparations for the use of facial recognition technology; first, the non-live aspect of it, and then the more retrospective aspect?

The first thing we will do is put some governance arrangements in place around any system we might procure to enable us to do that. On the other side, if I can talk about the benefit, the potential benefit of that is retrospective biometric recognition.

The Deputy will recall the violent disorder incident in Dublin in November 2023. There were over 20,000 hours of CCTV footage collected in the investigation and aftermath of those incidents. There was a team of 30 gardaí manually trawling through all that material, trying to identify the suspect, while also trying to rule out the material that was of no relevance to the investigation.

On the efficiencies that are to be gleaned from the deployment of retrospective biometric, for us from where I see it, that means 30 gardaí who could be in yellow jackets on O'Connell Street or around the city centre where we actually need them but they are inside. There is technology available and with proper safeguards, all the governance, engagement and guard-rails around how we would use that technology, I just do not see it as an efficient use of our resources when there is a machine that could do something much more efficiently as well. Even from a victim perspective, it is taking them months to trawl through all this material. That is where the benefit is.

That is the benefit but the downside is the danger of false positives, which have a big negative impact on those who are falsely identified, and false negatives, which means criminals can get off because the AI misses them for whatever reason. The evidence suggests the false positives and false negatives do not happen evenly across society but instead, as the committee has heard, the biases that exist within society are amplified by AI. For example, error rates are 60 times higher with west African women than with white men. What measures is the Garda taking to diminish or get rid of that altogether? Has the Garda taken into account the AI Advisory Council, which flagged substantial risks in the use of facial recognition technology?

Any deployment by us of such technology would be simply to categorise and to sort out material. There will always be a human in the loop. We will not be making any decisions or taking any executive actions arising from what the computer generates. There will always be that safeguard: the human will be there. We are also bound by the whole identification piece. It is not as simple as coming into court and saying, "The computer said it's Angela Willis" and that is accepted in court. We have to meet all of the evidential standards. Therefore, we have to have somebody who actually recognises that individual. They have to show how they recognise that individual and any previous interaction they may have had with them or else we leave them. It is not as simple as "The computer says". It is not like CSI programmes where the computer will identify who did it; it is much more complex. We are obviously accountable to the court for any decisions we make and there is a high level of proof required in that space.

I apologise for not being here earlier. I am attending another meeting at the same time but I have read both opening statements. I thank all the witnesses for being here. My first questions are for An Garda Síochána. How many cases are currently under investigation through the misuse of artificial intelligence?

Currently 224 referrals have been made to us that relate to intimate image abuse relating to children that we believe have been generated through AI.

Do the majority of those relate to the Grok example or does it span across a multitude of different platforms?

No, they relate to the Grok example.

Does the Garda have investigations under way at the moment outside of Grok?

Is that in relation to-----

The misuse of artificial intelligence in any way shape or form.

No.

Is Ms Willis aware of any complaints being made to An Garda Síochána about the misuse of people's images in other artificial intelligence tools?

In relation to adults, we have a small number. It is fewer than ten so I am not going to-----

Are there none relating to children?

In relation to children, there are 244 investigations.

That is in relation to that one case.

Yes.

That is fine.

Ms Willis emphasised that every algorithm must be governed by law and fully auditable. Can she tell us which AI systems is An Garda Síochána currently using or piloting and whether each meets the audit trail and transparency standards that she referenced?

I can because we are not deploying any at the moment.

Does the Garda have a pipeline? Does Ms Willis envisage that it will be using it in the short to medium term?

Probably not in the short term because we have governance and policy procedures to develop. So, it will not be in the immediate term but certainly in the medium term we would hope to be in a position to deploy some tools.

Would that be to help triage, say, harmful content online? What are Ms Willis's thoughts?

There are umpteen opportunities in law enforcement. We see that from our engagement internationally and with other law enforcement entities in the State. We see huge opportunities there. The referrals we get from the National Missing and Exploited Children Centre is an area where we see potential. I spoke earlier about a low risk introduction around some sort of knowledge hub where we would input all our policies and procedures, legislation, etc.

A lot of organisations are doing that, in fairness.

Yes.

Returning to the number of cases, I appreciate Ms Willis has given a number of cases that are being investigated by Grok but I am a little worried to hear the small number of other investigations outside Grok. I assume this is down to the lack of reporting or people coming forward with complaints to An Garda Síochána. Does Ms Willis think the gardaí should start some sort of information campaign to inform people of the complaints process available to them through An Garda Síochána when first, they feel their data has been misused or second, if their likeness has been misused? A lot of the Grok controversy was around the misuse of people’s images but there are instances the misuse of the image was done outside of Grok tool;– there are a lot of other tools there. It is worrying there has been so little reporting to the Garda. I am not criticising the Garda because it is not getting the reports and it can only investigate what comes to it. Has it looked at any type of information campaign to share with the public they way by which they can complain to An Garda Síochána?

The Deputy will be aware we work very closely with a number of partners but particularly with Hotline. Often the primary aim of the person impacted is to get the image taken down from social media, whatever platform it is on, and we engage with Hotline on that. Hotline makes referrals to us but only with the consent of the person who might have contacted it. We have engaged in campaigns but this is a societal issue. We can engage in further campaigns with our partners but I agree with the Deputy about making more people aware of how they can do this. It is something we plan to have more awareness campaigns on.

It would be worth working with the Data Protection Commissioner too, particularly on the misuse of people’s data and not just their images and likeness, which seems to be a lot of the focus of Hotline and the Grok controversy. I thank Ms Willis for that.

The PCSA stressed the need for heightened safeguards in policing. Does it have the statutory powers, technical expertise and resourcing to oversee high-risk AI systems that could be used by An Garda Síochána?

Yes, we do but we are not the only people in this space. There is also the Data Protection Commission, Fiosrú and the Independent Examiner of Security Legislation. Under section 234 of the 2014 Act, the independent examiner has specific powers because the Policing and Community Safety Authority oversees and assesses policing performance but when it comes to security that is an area we have no oversight over. The independent examiner has specific powers in relation to those areas.

If we were to look at international comparators, what oversight models that have been implemented in other countries would the PCSA consider best-in-class in developing our own approach to AI and policing?

The Deputy is assuming I have done in-depth research of other jurisdictions. As I said earlier, it is fortunate we were late to the party because other jurisdictions, in particular Canada, have made massive mistakes in this area and have been the subject of enormous controversy. They have undergone review and reform. As I said in reply to Deputy Gibney, Canada’s experience is one where the oversight agency was not involved in the discussions on the instruction of these technologies at the outset but came after to the party.

There is also a lesson we can learn from the New South Wales Law Enforcement Conduct Commission. We are very fortunate in the sense that we can learn from other jurisdictions. As the deputy commissioner outlined, we are at the infancy of the introduction of AI where in many jurisdictions it has become normal practice in the past five years.

I know we are tight on the clock, but my colleague wants to come in here.

The creation of the PCSA last year by the Oireachtas is actually a powerful change. There is really no best model out there. One of the big tools and two elements of our work will really bring this space forward. I mentioned earlier our focus on community safety and making sure the community are aware of when AI is being used and when it is not being used. The second piece is that we have inspection powers as well, and that is very different. I do not think there is another policing oversight body that has that combined inspectorate and public meetings piece. In the future, I see this as an area of focus in terms of some of the stuff the assistant commissioner mentioned earlier around governance. It will be our role to make sure that is happening on behalf of the public. That inspection role will be powerful in the future.

I thank the witnesses very much. I will start with the assistant commissioner. She mentioned earlier capacity in the cybercrime unit and the four regional hubs and the now small backlog. I will ask more broadly about resourcing and capacity in the system for pursuing cases of online abuse and child sexual abuse material being shared online. There are two issues. Is Ms Willis seeing a changing pattern increase in these cases and then in her resourcing to meet those?

In relation to having seen an increase, last year, we had 25,753 referrals from the National Center for Missing and Exploited Children. They are NCMEC referrals, so they are the companies, if you like, that come across this material, and there is a positive obligation on them to report that. That is done centrally through NCMEC, which is a US-based not-for-profit organisation. That is nearly a doubling from the previous year, so there is a huge increase there. We have had 2,450 referrals so far this year. Those come into the Garda National Protective Services Bureau. Then, I have another bureau, namely, the Garda National Cyber Crime Bureau. When we are involved in investigations of this type, we seize devices. That is where we talk about the backlog of examining those devices. Again, we see an increase in relation to the number of devices we are receiving but also the amount of child abuse imagery we are finding on those devices, so-----

That 25,000-plus is a huge number. It is a doubling from the year before. How is that processed? I presume they are not all the same thing. What triggers the bureau seizing devices?

It is a range of things arising from our investigations and information we might get internationally from international colleagues in this area. Then, we would carry out searches, etc. The Garda National Cyber Crime Bureau processed over 35,000 child sexual exploitation material files during 2025. Again, that capacity is there. However, if it continues to grow, this is where the use of AI will actually be of huge support to us. As I said earlier in reply to Deputy Murphy, there will always be a human in the loop when we are making a decision to take any further executive action or even just to catalogue and co-ordinate them. It could even something as simple as a European database of child abuse imagery. Our main focus is trying to identify victims and particularly victims we have not encountered before or who have not been encountered by law enforcement before in an effort to try to safeguard them. The tools would allow us to say we have already seen that image before and tried to safeguard the child or whatever. That would then let us concentrate on the new ones, if you like, where there is maybe a higher opportunity of actually safeguarding where that child is currently being exploited.

How many devices has An Garda Síochána seized? Has the number grown year on year? It is obviously a separate piece of work.

I am very happy to come back to the Deputy with the exact number, but I do not have it to hand. I do know it is constantly increasing. It is not only that the number of devices seized is increasing; the amount of data on those devices is just off the Richter scale.

Is the growth exponential?

It is quite significant because there is a digital element to every crime we investigate nowadays.

The nature of policing work has changed significantly.

Absolutely.

There is another thing that is relevant to this committee. Both in other committees and in the media, we have heard that these cases are slow. An Garda Síochána is often complimented on dealing with them compassionately, but they are slow and often go nowhere. How many cases go to court? How many prosecutions are secured? Are the numbers comparable with those for offline cases? Does this point towards a gap in legislation or regulation that the committee should be aware of?

Investigations are protracted because we are talking about multiple jurisdictions, different legal frameworks and having to obtain evidence from other jurisdictions that potentially operate wholly different judicial systems to ours. Some of them are judicially led. If we have to deal with another country, we have to engage in a mutual legal assistance process. I am not trying to justify delays or anything like that. We obviously try to expedite cases, particularly in this area. I draw attention to the fact that we recently signed a protocol with 15 other agencies in the criminal justice sector in Ireland, including Forensic Science Ireland, the courts, Tusla and various other agencies, to prioritise cases of rape and murder involving children. All of those agencies have committed to expedite and prioritise such cases and to have them deal with within a 12-month period. That is working really well. Complexity comes into it in where we have devices but must get the evidential piece from another jurisdiction. In our review, we are finding that is contributing to the delay. While all of the other entities will prioritise things-----

What about the pattern across cases in terms of the number that progress to court and the number of prosecutions that are secured? How do they compare? What does that data set tell us?

Again, there are complexities to these types of investigations. We are very reliant on victims. We have done a lot of work around our child specialist interviewers. We work very collaboratively with Tusla, the Child and Family Agency, in that space to make sure we minimise the impact on the child. I am very happy to send the Deputy some detail on prosecutions, conviction rates and all of that kind of stuff if he wants it but I do not have it to hand today. We bring as many cases as we possibly can before the courts.

Will the assistant commissioner to send us that information, particularly if there are patterns within it that might reveal legislative or regulatory gaps it would be important for the committee to be aware of?

I cannot remember which committee it was at, but I remember having an engagement on facial recognition with, if I am correct, the then Commissioner. During that exchange, I was wondering if both our understandings were wrong, if I was wrong or if he was wrong. I have tried to gather deeper questions to get to the heart of what it means and to undo the confusion I left the room with that day, especially given that the Garda Síochána (Recording Devices) (Amendment) Bill is before the select committee on justice, which I am not a member of, this afternoon.

We know it introduces a novel power around biometric analysis. This is not defined within the EU Act, but it gives An Garda Síochána the power to apply AI tools in respect of material such as imagery or video footage. We are told that this is not the facial recognition technology Bill as such, and we are told it is not about identification. However, it does allow gardaí to single out individuals who are known or not known to them. I have two examples of what I believe to be biometric analysis under section 43A(c) of the facial recognition technology Bill. In the context of the definition, I would love to be able to confirm whether that is also the witnesses’ understanding of how those examples would fit with biometric analysis. The first is where gardaí have CCTV footage of a suspect, person A, and they want to use technical means to compare images of that individual with publicly viewable social media photos and videos. In order to do so, they extract a biometric template of person A from the CCTV footage. This reference image is then compared with biometric templates extracted from a pool of social media images and videos. Would that meet the criteria of biometric analysis under the legislation?

First, we do not have the legislation yet, so it is very hard to say whether it would or it would not meet the criteria. I could not comment on that in the absence of the final legislation. I gave an example earlier to Deputy Murphy around the amount of effort and time we spend trawling manually through CCTV. The only reassurance I can give is that even if we drew comparisons from the material available to us, we would always need the human in the loop that I referred to earlier and a person who recognises the individual involved. We might use opportunities in the media. If there is someone we are looking to identify, we might use CCTV footage or a photograph of a person who is wanted for a particular crime on “Crimecall”, for example. Even when we get nominations for who that particular person is, there is a series of steps we have to go through before we knock on their door and ask them to come in.

If we are trying to provide examples of what we would define as use under section 43A(c), Ms Willis is saying that she does not know how that would apply to any given situation in terms of the use of biometric analysis.

Not until we see what the final legislation actually provides for.

I will go back to what Deputy Gibney said about the digital omnibus. Ms Willis said that whatever is in the legislation the gardaí will apply. I was thinking about the concerns around the GDPR within An Garda Síochána while we have the current legislative position in terms of the GDPR, which is much stronger now, prior to being deregulated.

I will give some examples where the use of data was questionable. One was the 1.5 million breath tests in respect of which bogus results resulted in wrongful traffic convictions. There was also the very sad situation in relation to Dara Quigley and the sharing of her image. The DPC found that gardaí in Duleek and Donore routinely failed to manually return the pan, tilt and zoom cameras to their original focus, so they remained focused on private residences. Thousands of 999 calls went unanswered. In 2022, the ICCL revealed that An Garda Síochána was unlawfully retaining files on innocent people who had been cleared of producing or sharing child sexual abuse images in the aftermath of it having mistaken images of children on beaches or in baths as child sexual abuse images. Most recently, An Garda Síochána lost 17,000 live-scan finger prints last year. If we are looking at deregulation of data, how can the public be fully confident considering the last few years? In the context of how data has not been fully protected in the last few years what has An Garda Síochána done to bring itself up to standard to make sure these situations do not happen again, never mind getting to the point of deregulation?

The Senator mentioned many incidents, some of which I am aware and others less so. I will speak about the 17,000 live scan records because that issue falls within my area of responsibility. There is no such thing as a loss of 17,000 live scan records. Work is ongoing in this regard and I am confident that when we get to the end of that piece of work, we certainly will not be talking about 17,000 lost life scan records. It seems to have been due to an anomaly between two ICT systems not properly communicating with each other. I am happy to come back to the Senator in this regard when we get to the end of that work.

I am not familiar with the road traffic piece the Senator mentioned. I am not sure whether the Senator has any further detail in that regard.

It was in the context of breath tests. The question is more to do with the GDPR and data breaches that are all in the public realm. Before this deregulation happens, I wish to know how An Garda Síochána has sought to address the technical data breaches that happened under current stronger legislation. I am not asking Ms Willis to respond to individual cases. Rather, I seek to know what An Garda Síochána has done over the last few years to improve its data protection.

We have a layered approach to access to information. We are mindful of the vast amount of information we hold on members of the public. We have an individual sign-on system to all of our systems, which leaves a sort of a fingerprint of anyone who accesses any of the information we have. There are significant governance measures in this regard. I can let Mr. Fintan Brady speak in a minute, if the Senator so wishes, about cybersecurity and the layers of governance we have around the information we hold and how we protect and safeguard that information from falling into the hands of people who are not entitled to it. We have our policies, procedures and training of our people, as well as that governance piece around access to the information we store.

When it comes to the deployment of any AI technology, as I said to the Senator, we will always ensure there is a human in the loop. We will, therefore, be able to go back and find out who the human in the loop is and exact accountability. We have internal accountability and then we have a range, as has been alluded to earlier, of external oversight that holds us to account on how we do our business. That includes how we process information.

I thank the witnesses for coming before the committee today. I will ask some questions around the existing use of technology within the force and the potential for AI to support operational improvements or efficiencies. I will start with the interview rooms. Today, as things stand, if a suspect or, indeed, victim of a crime is being brought in to give a statement, am I correct that there is audiovisual technology in the interview rooms?

Yes.

Am I also correct that such information is not relied upon in a court case, for transcription or as a reference point for statements taken?

There is an obligation on An Garda Síochána and whomever is undertaking that interview to write a summary of the content. Obviously, it is not possible to write a contemporaneous note of every word uttered. A contemporaneous note of what went on is taken and that is relied on in the courts. On occasion, the tape will be played before a trial, jury and court.

In the usual course of an interview or statement being taken, notetaking is happening in the room by a garda. There will be two gardaí present. In that situation, the statement has to be read back to the person giving it before the interview can be concluded.

Yes. They are asked whether they are satisfied with the content and summary and then they are asked to sign it. Sometimes, they will sign it and sometimes, they will not.

That has to happen in real time.

Yes.

In that instance, would AI be potentially a tool used in a closed circuit situation where it could be assisting in transcribing those notes and assisting An Garda is using more of their time in the questioning and interviewing of the person in the room, as opposed to the notetaking?

Yes. As the Senator will be aware, the Garda Síochána (Powers) Bill is currently being considered as well. I will not use the word "lobbied", but this is something that we have advocated for for years to remove the obligation to write down. It interrupts the free-flowing conversation in the interview. Within the Garda Síochána (Powers) Bill, we anticipate that the writing down by the member may be removed. In that event, we will have to move to technology that will actually record and also transcribe and hopefully translate as well. The Senator will be aware that many of our interviews require interpreters if English is not the person's first language. We see huge opportunity in that, not only that it would give a contemporaneous note, obviously we still have to go back over and make sure that the system is doing what it is supposed to be doing and transcribing, but we would see a scenario where there might be a screen and that it would be transcribed as a person is talking in real time. We would still go back over it. It is not only the writing down in the interview, we then have to give that clerical support and they have to type it all out. There is huge opportunity in that area. That could be made available electronically to the DPP, etc.

In terms of report generation, as I understand it, a considerable amount of work is involved for the investigator of a crime in preparing all of the evidence in the documents that are required before a file can be sent to the DPP. In summarising that alone, there is a significant document required, a summary or cover document. Am I right in saying that?

Yes.

It can extend to 40 or 50 pages depending on the crime. Am I also correct in thinking that a sergeant has to give a synopsis for the same pile as part of it? It is not just one person's time that goes into the administration around summarising and synopsising the case. A sergeant or an inspector would have to give a covering report that would be going with those files.

It depends on what the incident is. If it is something as simple as shoplifting, it can be a very neat, short file. A murder or organised crime file that is going to the DPP will generally have what we call incident room co-ordinators. They are essentially a trained member of An Garda Síochána. They assimilate all of the information, and then they will compile the report the Senator referred to. On top of that, the superintendent will formally send that report on to the DPP. Within that, there might be a high-level summary of what is contained. It would not be verbatim. We will have one detailed report, which will direct the directing officer to a particular statement that might have what we would consider relevant evidence-----

The roadmap to the case.

Continuity and all of that stuff is covered in it. A high-level overview is then provided by the superintendent.

Presumably a considerable amount of time, attention and detail is required - and correctly so - for that. However, in a world of AI supporting word processing or report generation, again, not in an open-sourced system but in a closed system that is unique to An Garda Síochána, does the assistant commissioner see opportunities for AI to be deployed to support the report generation in the first instance - draft report generation?

Potentially. Again, it depends on where you are storing that information. We are a very long way away from actually uploading statements to any system that would interpret them and provide an overview report.

I understand that, but I am asking about the covering document that states where each statement and position is - the general, boring, tedious word document work.

It is available, but I would not see it as a direction in which we will be going any time soon, but maybe in the medium term. Again, it would only be under very strict rules, supervision and all the rest of it.

Very secure, it would have to appreciate the sensitivity of the information and personal data and everything. For people not in the assistant commissioner's line of work, those kind of tools are used on a regular basis to assist with administrative work.

I think most people would understand how a certain amount of the administrative burden could be relieved with using AI.

The opportunities are endless in this space. However, we are obviously constrained by certain regulatory and administrative------

I would like to ask about what happens when a complaint is received from an individual - an adult - of harassment, threats or non-consensual sharing of intimate imagery. When someone comes into a regular Garda station and presents that information on their phone, does the capability currently exist in the hardware in a Garda station for a member of the force to look up that social media site and review that independently themselves?

That would be sent to one of our cyberhubs and they would have that capability.

Am I right in thinking that access to social media sites is restricted in most Garda stations?

All of our members have a mobility device, so a mobile phone that has access. I will turn over to Mr. Brady. We have rules around what can be assessed and the reasons it can be accessed.

Just briefly.

The access we can provide to people on that is very granular. It is not so much a geographical location that might have access; it is more in terms of the role the member of the force may have. It is entirely feasible to grant them access to those kinds of systems if it is required for the job but in the mainstream it would not be available to people.

Okay, so-----

We will have time for a second round.

-----in that instance, the garda may have to take the person's device, hold it as evidence and send it to the cyber unit.

As the assistant commissioner has mentioned, it may be transferred to a cybercrime hub.

We will have time for a brief second round.

I thank all the witnesses for their statements and for answering a wide range of questions in detail. I will start first with the assistant commissioner. I want to ask a number of questions. In recent years we have seen a rise in scams in Ireland and financial scamming. A lot of it is probably happening from overseas as well and is targeting Ireland. Probably everybody in this room has been subject to some form of a scamming attempt via technology. I wonder if AI is contributing to this and if there are current investigations where AI has been used in scams. Does An Garda Síochána feel there is a need for a stronger public awareness campaign in relation to that?

Yes. There is no doubt that AI is contributing to an exponential increase in online scams and all sorts of online behaviour from deepfakes to all the phishing and smishing text messages and phone calls. We talked earlier about this evolving so rapidly. A short snippet of your voice can then translate into a demand for money or whatever else and that can be targeted at your family member who might think they are talking to you. This is certainly an area about which we need to further inform the public. We do lots of information campaigns around trying to prevent people, particularly vulnerable people, from getting hooked into these scams. We did one recently around St. Valentine's Day and romance scams. That is one of the highest areas where people are being targeted. Again, vulnerability is being targeted. People do not have the courage to come forward and report that because they feel ashamed that they have been caught out in this manner. It is an area in which we constantly try to do more campaigns but we need to do more, and collaboratively with the other people in this space as well. There are other agencies that we co-operate with, such as hotline.ie or Cybersafe Kids if it is relation to children.

The assistant commissioner mentioned the romance scams. To follow up from that, would An Garda Síochána have much involvement with dating apps in relation to scamming or have there been any interactions there?

We engage with the companies, the platforms, but not really with the people behind the apps. Generally, they are in other jurisdictions. They are not located in Ireland. We certainly engage with the platforms around this to see what kind of prevention measures we can do and how we can work together.

Regarding recruitment within An Garda Síochána, does the assistant commissioner feel there is a need for further recruitment in relation to upskilling when it comes to AI and technology in general?

Has the PCSA advised on any training in relation to this? It was mentioned that the PCSA oversees the proportionate use of evolving technologies. Have there been any incidences where it felt it needed to advise limiting the use of evolving technologies within the Garda in certain areas? Has that ever happened?

As the assistant commissioner outlined, we are very much at the infancy stage. We have not had recourse yet to evoke those principles. Deputy O'Rourke said something about child sex abuse imagery. I note from our monthly meetings with the Garda leadership team that a welfare piece is there as well, not just a technology piece. Some of those images are of the most extreme violent nature you can imagine. To have gardaí looking at those day in, day out, is very difficult. In my other role as a barrister and seeing these cases before the courts, it is something that is unimaginable. Technology that would assist gardaí in not having to look at those until the end product would be something that would be very beneficial. I do not think we can underestimate the harm, not just to the victim and the child at the centre of those images but to the people who are prosecuting in respect of them.

In terms of us advising the Garda on recruitment, I will bring in my colleague, Ms Hall.

I thank the Senator. Over the years, even the predecessor body to the PCSA had significant recommendations around making sure the organisation and the police service are appropriately resourced. I mean that in the broadest sense. Her question is very apt. That investment in technology expertise is expensive in the market in terms of getting people involved. I am sure Mr. Brady will have views on this as well, as will the assistant commissioner. For a number of years, we have said there needs to be significant investment. Some of that is not necessarily Garda members. In fact, it probably is not. It could be in the area of AI specialists, IT specialists and human rights specialists, who can look at the use of this and be part of the governance framework for any of this emerging tech. It would not necessarily be us.

It is an interesting role. Our oversight tends to be us making a report on something we thought was lacking. "Civilianisation" is the wrong word, but it is about professionalisation and making sure the likes of the assistant commissioner investigation has the appropriate people to support the investigation. They would not necessarily be people in uniforms or not just clerical people either. There is a huge change in getting that. Some inroads have been made in An Garda Síochána but there is a lot more to be done, both physical and personnel, in terms of that expertise that will need to be done.

I will add to that. We are trying to upskill and hire specialists in this area, and people who can assist us with forensic analysis, etc. Our challenge is that this is a very competitive market at the minute because it is so new and emerging and goes across many areas in both the public and private sectors. It is about retaining the people we hire and trying to hire them at the appropriate grade we are allowed to pay them at. They are snapped up by private companies. They are attracted out with much greater financial packages than we can offer them. That is certainly something that is a huge challenge for us.

It falls to me to thank our witnesses. I very much agree with Dr. Byrne on content moderation. This issue came up before the media committee. It may be appropriate to deploy AI for content moderation in those circumstances, from a welfare perspective. As Senator Ryan mentioned, its role in assisting with statements is something that can be explored. It is always critical that there is a human in the loop, but there are certain examples where AI can be used.

I will come back to the issue of scams. The technology at the moment is such, and I have had this done, that a voice can be taken in a matter of seconds, the tone and tenor can be picked up and a speech can be generated.

That is going to have implications. People are used to getting a dodgy text or email but it is completely different when they hear somebody's voice or see their image. There was a very well-publicised case of a woman who thought she had engaged in romance with Coldplay's Chris Martin. It was extensively covered on RTÉ. When the assistant commissioner was responding to Deputy Ó Cearúil about the numbers of those kinds of cases, she indicated that fewer than ten had been reported to An Garda Síochána.

There were fewer than ten that involved AI-generated intimate image abuse of adults.

They involved intimate images. However, in terms of what I would call deepfake scams, where you have people who believed AI-generated materials, for example, we regularly see on our social media feeds well-known personalities advising on financial advice, how many of those type of cases are there?

I am not aware of any of those cases that are under investigation.

Civil actions have been taken. I ask primarily because I have a Bill on this issue. There is the offence of fraud but does Ms Willis believe there is a criminal offence in the generation of the deepfake with the intention to commit fraud?

If it constitutes a crime and is currently on the Statute Book, we will absolutely investigate it. It generally is done with the intent to commit a fraud or some sort of a crime or deception and to trick the person into believing that it is a real image, voice or whatever it is that is being manipulated.

An Garda Síochána's responsibility includes dealing with organised crime, the perpetrators of which are often particularly smart in trying to deploy new technology, particularly, as Ms Willis mentioned, personalised phishing or synthesised identities. Is Ms Willis starting to notice any of that in her engagement with other jurisdictions?

We see it particularly in the economic crime areas. A criminal network will send out perhaps a million texts. It only needs a few people to engage with that to achieve its aim. The criminals have the technology available to them at their fingertips, perhaps in their bedrooms, in a train station or wherever they are. They just send it out and it is there. These have been attempted on all of us on numerous occasions. It goes back to the piece we talked about on informing the public as the technology evolves. The Cathaoirleach spoke about nearly having a password if somebody is ringing him so that he knows it is a genuine call. We are nearly getting into that space because the technology is moving so rapidly.

We have talked about family code words given the potential of AI-generated voice scams. Are there legislative or policy areas the assistant commissioner thinks will be able to strengthen her hand as to how we can address some of these challenges?

One of the greatest challenges we have is turning the behaviour into evidence that we can bring before the court and that is acceptable in a court setting. Anything that can be done to make that a smoother and an easier process would be welcome. As I said, sometimes that information can be in another jurisdiction, which brings its own challenges for us. Again, we deal with whatever we can within the legislative framework we have and any strengthening of that would certainly be welcome.

Forgive me for teasing this out because there is obviously a difficulty. Where there are jurisdictions with which we have a good relationship - a lot of democracies and so on - it is relatively easy for An Garda Síochána to engage. Part of the challenge is that some of this data is stored in jurisdictions which, for whatever reason, may not have as good a relationship with us. What is An Garda Síochána's experience in that area and what more can we do?

We get good co-operation from most of the platforms when we seek the information. The first step in that process is that we put a preservation order on the material and then we can go through the legal process to get it evidentially.

We have good engagement and co-operation by and large. However, if we are getting into the space where we are asking a platform to incriminate itself, we are not sure how that will play out. Perhaps something could be done in that area. Then, the other piece-----

Will Ms Willis be specific? The role of this committee is to make recommendations. What specifically does she think would strengthen her hand with regard to the platforms?

It is quite complicated because we are trying to validate the information we are looking at and the intent of the individual to commit the crime, whether it is fraud, deception or whatever else it may be. An easier evidential process for us would be better. I know we were going to talk about mobile devices. Sometimes we need the actual mobile device. If it is a significant case involving a lot of harm, we will retain the device. If we could download it and a copy of that device becomes acceptable within that criminal process in the same way as the original, that would be helpful. Sometimes we are only looking for a tiny piece of the information that is on a device but we end up taking the device from the individual. Their whole life can be on that device. That is really harmful for the victim.

To follow up on that, the late Assistant Commissioner John O'Driscoll was a very strong proponent of what the assistant commissioner just advocated. What happens in domestic violence situations is that there is pertinent material on the mobile phone but in order for that investigation to continue, the phone has to be taken off that person. That phone, essentially, is the person's life. It will have all their contact details. People have everything on their phone. The late Assistant Commissioner O'Driscoll proposed that victims of domestic violence, for example, would be given a new mobile phone. If we had this technology, it would allow some material to be taken off and then the person could continue using the phone. I am fearful that there are people not progressing very valid complaints because they need their phone for their work and whatever else.

That is worthy of consideration. We have time for a quick second round. It will be three minutes.

I thank the witnesses for staying. For clarity, the assistant commissioner said there were 244 investigations under way. Were they into Grok? Is that the point she was making when Deputy Murphy asked the question?

Sorry, those were 244 referrals through NCMEC which related to AI-generated material.

Ms. Willis clarified that Grok was the sole-----

Yes, in relation to that 244.

She said that 50% of those relate to child sexual abuse imagery.

Yes.

There are, therefore, about 120 investigations into child sexual abuse imagery.

Yes.

Was An Garda Síochána provided or has it procured images as part of these investigations?

Yes. That would be part of the process.

Has it yet been established if any of those images constitute or may constitute child pornography within the meaning of the 1998 Act?

About half of them, 50%, yes.

What An Garda Síochána is investigating now is presumably the person who produced that imagery. Can Ms. Willis say with confidence that those images that are in the possession of An Garda Síochána were generated through an AI system?

We do not have any technology to assist us in that determination. We are relying on the experts who work in that area in the online child exploitation office but we are reasonably confident that the images are AI-generated.

Has An Garda Síochána had any interaction with counterparts internationally which may be carrying out investigations similar to those it is carrying out?

Yes. We engage through Interpol and we know that other jurisdictions are similarly investigating this.

Is the assistant commissioner concerned that existing AI systems that are accessible to people from within this jurisdiction could still be used for the purposes of generating child sexual abuse imagery or child pornography within the meaning of the Act?

It is possible. It is important to say it is not just Grok. There are many nudification apps available.

However, it is the case the investigations the Garda is dealing with solely relate to Grok.

Yes.

What can we do about this right now if it is the case AI systems are being used to generate child pornography? That is a pretty alarming vista we are facing. Ms Willis's job is to investigate current crimes but what can the Minister or the Government do to respond to this reality?

We have had extensive engagement with Coimisiún na Meán on this. There is a regulatory piece and the initiation by the EU of an investigation is a major step in that. Obviously Ireland will co-operate with that investigation around the platform. We are investigating what has come to us and we will take that as far as we can from an evidential perspective. It is that regulatory piece around actually allowing this to happen in the first place because by the time it comes to us the harm has been caused because the image has been generated and circulated. Our members have to view that material then, as the Chair said, and that can have a harmful impact on them.

I am going to ask about AI and the use of it in the committing of crimes whereas previously I was looking more at the use of AI in the Garda itself. I am interested in scams. The assistant commissioner answered some questions on this already but perhaps we can delve a little deeper. I am specifically interested in pig butchering scams. Is Ms Willis familiar with that term?

I am sorry.

They are commonly known as pig butchering scams. Essentially, I think is the idea is that somebody is fattened, so to speak, through relationship-building until their wealth is siphoned from them and often their surrounding communities over the course of three to four months. Is Ms Willis familiar with the trends in that space in Ireland? According to research from the University of Texas, between 2020 and 2024 $75 billion in global wealth was transferred through these kinds of scams and $5.7 billion alone from the US in 2024. Does the Garda have those kinds of figures for Ireland? How much wealth was scammed from Irish citizens in, say, the last year?

I do not have those figures but I can come back to the committee if we have any information that might be useful. Again, we would only be estimating the amount of money in this space, apart from any crimes we might be investigating in this particular category. I am very happy to come back to the committee with some further information in relation to that.

Please. I assume this is handled by the Garda National Cyber Crime Bureau.

If it relates to fraudulent scams, it would be the Garda National Economic Crime Bureau.

How many members are in that bureau?

There are about 120 in the bureau.

There are about 80 in the cybercrime bureau.

Yes.

And about 14,000 in the force.

There are 14,300 at the moment.

So we are looking at less than 1% dealing with these types of crimes. Does Ms Willis think that is enough? Does she think we need to grow?

I think this is an area of growth. There are lots of competing priorities for the Commissioner at present and obviously front-line, high-visibility policing is one and security is another, as we have the EU Presidency coming. There are lots of competing demands and obviously the more resources we have, the more we can do in this space.

Coming back to the scams specifically, has the Garda ever investigated scam farms here? The places where these scams are done are usually located in parts of Asia, at the moment primarily Myanmar and elsewhere. However, a few years ago one was uncovered on the Isle of Wight. Has the Garda ever had any suspicious activity that relates to this in Ireland?

We have no information to indicate there is one in Ireland. We certainly collaborate with our European counterparts and we have people here connected but not actually directing it, if you like, from here.

Chair, we might look at Singapore because it is pretty much the only country that is starting to raise awareness of this issue through public campaigns. It would be worthwhile to look at what it is doing.

In relation to grooming and CSAM, I am wondering about the Article 27 countries and companies that are not based here. Alternatively we could take Roblox, which has AI capacity within it. We also had the "Prime Time" programme and I think Roblox's EU base is in the Netherlands.

There are the 244 cases looking at Grok, and X is based here, but has the Garda enough resources or is there good enough co-operation with other jurisdictions, for example Article 27 companies that are based in the US or companies that our teenagers are using but which are based in the Netherlands?

We have good co-operation with the platforms regardless of where they are based. As I said, that first step in the process is to serve what is known as a preservation order so they will hold the material and then we will look for it afterwards. On the resourcing, it might be of interest to the committee that in 2025 between the Garda National Cyber Crime Bureau and the Garda National Protective Services Bureau - there are two sides when we are talking about online child exploitation - there were 161 identified and safeguarded between those two bureaux during 2025. Clearly, the more resources we have to work in this space, the better. It is one of the highest-harm areas we deal with. The more resources we have in that space, the more we can do.

Of the 161, does the Garda have data on who the perpetrators of the abuse were? Were they mostly based in Ireland or outside the jurisdiction?

It is a combination of both. Our primary aim in this space is to safeguard the child and get the child to safety. There were 16 of those children located in Ireland. With the other children, we are engaging with our counterparts in other jurisdictions and we are trying to safeguard the child first and foremost. Then the investigation will follow in relation to the perpetrator. Sometimes that perpetrator might be present with the child when we are safeguarding the child.

I want to follow up on the point about the scams. One of the scams that was very prevalent a couple of years ago, though I have not seen it in a while, involved the use of prominent people like Ryan Tubridy, Micheál Martin, Roderic O'Gorman, Miriam O'Callaghan and Bríd Smith in a video where they say that they were caught out and that there is this thing you really need to invest in, if only you knew about it. It is using people's likenesses to advertise a product that is a financial scam. I have seen there have been private cases taken to bring this stuff down but have there been any criminal prosecutions or investigations into any of these scams?

I am not aware of any but again I am happy to confirm that and come back to the Deputy.

Okay. Related to that, on the question of misinformation and disinformation, have there been any Garda investigations into that? I can imagine it very seriously impacting on our democracy. There will be a general election in some years' time and in my constituency, say, you could have a video from me or which seems exactly like me saying in this area people should vote number 1 for someone who is not me as that is the best way to get a left government. That could have a very serious impact on the outcome of elections. Is something like that a crime? What is it a crime under and have there been investigations into things like that?

The Deputy is talking about an impersonation-type scenario.

Exactly.

I am not aware of any prosecutions in relation to that but we certainly have had investigations into misinformation and disinformation where real people, if you like, have put information online encouraging people to commit crime in certain scenarios in the city.

What about stuff that impacts on the running of our democracy? What are the charges the Garda can use there if people are spreading information in a very malicious way? Again, I think during the presidential election there was some AI-generated stuff of Catherine Connolly. Presumably this is a crime but what is it a crime under?

You have things like incitement to hatred. It depends on what the content is. Every situation has to be examined depending on what the material we are looking at is. Incitement to hatred is one of the main ones that comes to mind.

What if there is no incitement to hatred or group targeted? Let us say there is a false video of me saying the best thing to do is not to vote for me but to vote for someone else? Is it a crime to create a false video of me?

If a complaint is made, we will investigate it.

Does Ms Willis know what legislation it is potentially a crime under?

Potentially, it is an impersonation offence. We would have to see what the situation was, talk to the Deputy, get a statement of complaint from him and see where we could take it from there.

Thanks a lot.

I would like to go back to section 43A(c) in the recording devices Bill. Ms Hall referred to how the public understands the use of artificial intelligence and how important that is. I read the section again. Given that the Bill is Government legislation and already on Committee Stage, the likelihood that its current structure will be its final structure is very real. I am wondering about the fact that we cannot seem to produce a step-by-step outline of how section 43A(c) would apply. If we do not have clarity on it, how can we think about how the public would understands it? Could somebody give me a step-by-step outline as to how the recording devices Bill, specifically the section on biometric analysis, will be used?

It is not used at the moment because we do have-----

I am referring to the Bill that is on Committee Stage.

When the legislation is enacted, we will examine it. We will have to put a policy and procedure document in place in respect of it. As part of that process, we will engage and do our human rights screening of-----

Has the Garda had no engagement with the Department of justice on legislation that would give it a novel power? This is in the context of not being able to give scenarios as to how the legislation will be used and of the possibly of deciding how it will be used retrospectively. It feels like it is going in the wrong direction.

We do not have it at the minute, so it is a matter of trying to predict what it is going to look like.

The Bill is on Committee Stage. It is a Government Bill, so it is most likely to be passed. How do we explain it publicly if we cannot explain it to ourselves? The detail, including that relating to section 43A, is available, and we can see how things will be defined. At this pint, we should be able to say how it would apply to those given powers under it. I am concerned that we cannot communicate that. It is concerning for all of us, especially if we are to communicate to the public how the legislation will be used.

When we start to implement that legislation, we will engage with the public and tell them how we are going to proceed. We will have a policy and procedure. The policy will be available publicly. The vast majority of our policies, bar a few sensitive ones we have in various areas, are available publicly. I envisage consultation on how we will implement the legislation. I am sure we will be subject to examination at our oversight meetings every month. We are very much in the transparency mode and we are not going to be doing anything behind-----

I am even concerned, from a political perspective, that politicians will allow legislation on which they do not have clarity to go through. Even before it gets to the point of being applied, the understanding of what it means seems to be quite wishy-washy.

I will return to my questions on the current use of technology within the Garda and the opportunities for improving this as a support to members of the force. In what areas does the assistant commissioner envisage enhanced use of technology to support members of the force in the performance of their duties, given that there is huge demand for more visible policing and appreciating that there are, commonly, social media and digital aspects to many crimes?

In our crime investigations, I see huge opportunities. We talked about the area of child sexual exploitation and the vast amount of information we have to trawl through. That is certainly an area. We talked about online fraud. Again, that is another area. We have to automate many of the processes in the background and undertake analysis in the background. Again, there has to be a human at the other end of it if we are going to make any decisions around what further steps we might take. With the introduction of the Garda powers Bill, we will need to redo all of our interview rooms, interview suites and child specialist interview suites. The translation and transcription service would be very beneficial. There are many other areas. Maybe Mr. Brady might like to refer to our financial area and HR filing systems. It is about making all of these much more streamlined and accessible to users internally.

I asked the question because I appreciate that there is probably strong awareness among the public of the need to digitise our healthcare system, but it strikes me from listening to certain of the answers today and from interactions with members of the force that there is a huge need for us to deploy a digitisation strategy or improve the plans under way in that area, and accelerate delivery to optimise the use of existing technology, let alone new technology.

Exactly. As part of our transformation programme, we are on a journey. The investigation management system is one aspect. Historically, all of our investigations were paper based. Files were stored in filing cabinets or put away in boxes. All of that information is being digitised now, and that means it will be available forever, which it needs to be.

We are on a journey. We introduced body-worn cameras, and we recently introduced Tasers. There are different projects that we are implementing. There is a limit to the amount we can absorb at any given time because all the transformation, which involves training and new policies, means there is a lot coming at our members. They still have to do the day job. Therefore, it is about trying to sequence everything in order that it can be absorbed and that we can still be efficient along the way.

Representatives from Comisiún na Meán were before us last week. I asked them if the organisation is monitoring the dark web. It is not. Given the prevalence of child sexual exploitation material and images created through artificial intelligence, is the Garda monitoring the dark web?

We are not actively monitoring the dark web.

When evidence comes before the Garda or there is reason to believe that there is either misuse or there is material on certain parts of the dark web, it is under those circumstances that it investigates?

Yes.

Does it have the technical expertise to do so?

Yes, we do.

Why is there no extensive monitoring of the dark web? Why it is monitored only when a complaint arises?

It is down to capacity. It is as simple as that. Our capacity to monitor it is limited. I am not sure whether the Deputy has ever ventured there, but it is not a very-----

It is really a capacity issue for An Garda Síochána. If we were to deploy AI tools in that space, there would certainly be huge opportunities for us.

Has the Garda assessed any AI tools to help it regulate the dark web and images on it?

No, we have not assessed any tools to date.

I was shocked to learn last week that there is no active monitoring of it, particularly considering that is where the worst type of material, as Ms Willis alluded to, predominantly resides on the dark web. It is not accessible on the Internet to 99.9% of people. It is on that dark web. There needs to be more of an emphasis, from a regulatory point of view and from An Garda Síochána, not only on acting on complaints related to the dark web but actually monitoring it actively and getting ahead of particular issues.

I address the same question to the Policing and Community Safety Authority. Is this an area it has looked at as it relates to the duties of An Garda Síochána and how it would be monitored in terms of how it uses and monitors the dark web?

The simple answer is "No". We do not ever get involved in the investigation of individual cases. No, it is not something we actively looked at.

I appreciate that, but I am asking about oversight of An Garda Síochána in terms of how it investigates.

At times it is more about looking at whether the Garda has the expertise to do it is doing and whether it has the systems in place. The assistant commissioner mentioned that over the past ten years, An Garda Síochána has moved from literally having big, physical journals for their most serious crimes to now having a technological solution in terms of the investigation management system. The Garda is probably light years away from contemplating the AI element and an awful lot of investment will be needed, even to bring it to that point. Active monitoring of something as vast as the dark web is not something we are really looking into at this stage.

I thank the witnesses for their contributions. We should see technology as an enabler. It is not a goal in and of itself. If a technology such as AI can be used to help solve crime, guarantee national security and so on, then we need to use it in those circumstances. To follow on from Senator Ryan's point, I can see that we would be able to use AI to help with the administration of Garda stations, in areas where there is a large amount of data involved, financial crimes and so on, and when dealing with content moderation. Our challenge is that we want to ensure there are humans in the loop and that there is human oversight. This is all going to come at us very quickly.

Dr. Byrne mentioned that the adoption of AI by An Garda Síochána has commenced. While I appreciate the point made on investment, I am pretty certain that within the next few years, An Garda Síochána will be using AI and there will be pressure to use it and other technologies.

What oversight should we be looking at and what kind of safeguards do we need to put in place with regard to policing and justice matters? That is a question for the Policing and Community Safety Authority.