I advise members of the constitutional requirement that they must be physically present within the confines of the Leinster House complex in order to participate in public meetings. The evidence of witnesses physically present or who give evidence from within the parliamentary precincts is protected by absolute privilege in respect of the presentations they make to this committee. However, they are expected not to abuse this privilege and it my duty as Chair to ensure that this privilege is not abused. If their statements are potentially defamatory in relation to an identifiable person or entity, I will direct them to discontinue their remarks. It is imperative that they comply with any such direction.

Members are reminded of the long-standing parliamentary practice that they should not comment on, criticise or make charges against a person outside the Houses or an official either by name or in such a way as to make him or her identifiable.

Our meeting today is the first in a series of meetings on the topic of artificial intelligence and the environment. It is a very important topic and I am glad we can discuss it. We are pleased to be joined by Professor Andrew Parnell, Met Éireann professor of data science for climate and weather and director of the AIMSIR research institute at University College Dublin, UCD; Professor Hannah Daly, professor in sustainable energy and energy systems modelling at University College Cork; and Professor Jennie C. Stephens, professor of climate justice, member of the ICARUS Climate Research Centre at Maynooth University and co-convener of the Climate Justice Universities Union. I thank the witnesses for coming to share their insights. Each of the witnesses will make an opening statement.

I thank the committee for inviting me. I am the professor of data science for weather and climate at UCD. I have several roles that might be of interest to the committee. I am the director of the Research Ireland innovate for Ireland centre in AI for decarbonisation and the deputy director of the Research Ireland co-centre climate+ for climate, biodiversity and water. Both of my roles in these research centres cover the intersection of AI and its application in mitigating and adapting to the climate and biodiversity crises on our island. In particular, I focus on training the next generation of weather and climate scientists to be fully versed in the foundations and applications of state-of-the-art AI.

My main role is as director of the AIMSIR research centre. This research centre is funded by Met Éireann and looks at how AI can be used to improve our national weather forecasts. It started in September 2025 and funds around 20 projects looking at areas such as weather extremes, short-term rainfall forecasting, floods, storms, power outages and many other topics. Each project is led by a set of UCD AI and domain experts and Met Éireann staff.

The world of weather forecasting has been completely upended by AI over the past three to four years. Our traditional weather forecasts are based on mathematical and physical equations of the atmospheric system and have served us well for the past 70 years. For example, back in the 1960s, our error in estimating the track of an Atlantic storm for where it would be in three days' time was over 700 km. By the early 2020s, with satellites and our modern computational systems, it has fallen to around 100 km. The one-day forecast is now below 50 km. These advances have saved countless lives both in Ireland and across the world.

However, those figures are all based on our previous mathematics- and physics-based approach to weather forecasting. In 2022, the European Centre for Medium Range Weather Forecasts, funded by 23 European countries, including Ireland, released its AI weather forecasting system, AIFS. Depending on the metric used, it beats the latest traditional weather forecasting approaches by 10-15%. That is roughly the equivalent of a decade of research in a couple of years.

It has been in operation since 2025. Across Europe, it contributes to the daily weather forecasts we see on TV and on our phones. Perhaps the most remarkable statistic about these new AI methods is that they are estimated to be 1,000-times more efficient at producing a weather forecast than our previous physics-based approaches. AI weather forecasts can be obtained in a fraction of a second even on a standard laptop such as members might have in front of them right now. This has implications both for the emissions associated with the AI model runs, which are much more efficient than the traditional ones, and for the democratisation of weather forecasts because they are no longer solely in the hands of rich nations that can afford the computational power to produce them.

Another huge advantage of some of these new AI weather forecasting models, such as the AIFS method that I mentioned, is that it is open source. This means that the data used to train the model is entirely open and available to everybody. The model weights the structure available on the web and anyone who has the right level of computing can train a new weather forecasting model. Indeed, the whole field of AI-based weather forecasting started off just a few years ago, in 2021, by one person who attempted to apply some new AI methods to some data that was available online.

It is important to note the difference between this type of AI, which, although based on many very similar mathematical concepts and computational techniques that can be found in large language models, LLMs, is very different from them. The datasets are far smaller. The emissions footprint, even during the training of the AI model, is tiny compared to a large language model and smaller than what is already being used to create a traditional weather forecast. These models fall outside the remit of the EU AI Act.

We might refer to this type of AI as data-driven AI or machine learning rather than language-driven LLMs. Ireland has excelled at data-driven AI over the past few decades through initiatives such as the Research Ireland research centres and the centres for research training. These Government-funded AI-related initiatives are something I really hope that this committee can continue to support and encourage.

I realise that the committee has heard a lot of negative aspects of AI around issues to do with security, data sources, defence and use of natural resources. However, this particular area of data-driven AI models is a topic that still needs support from members and from the public at large. We have a wonderful education system that is producing high-skilled graduates in these areas, but we need more support in those AI models that have a net benefit to society and should not be included in the same category as frontier LLMs. Aside from the training that my centre offers, the data sources through which Met Éireann contributes are enormously important for international weather forecasting and must continue to be supported and expanded if we are to reduce the uncertainty in our weather forecasts further. Purchase and deployment of the new computation, analysis and simulation platform for Ireland, CASPIr, high-performance computing system through the Irish Centre for High-End Computing is also vital for keeping Ireland at the forefront of the data-driven prediction aspect of AI.

I thank members for listening.

I thank the committee for the invitation to appear today. My research focuses on energy systems, decarbonisation pathways, carbon budgets and climate policy. Over the years, I have undertaken research funded by Departments, the Environmental Protection Agency, EPA, the Climate Change Advisory Council, CCAC, and a range of civil society organisations. My research has played an underpinning role in the evidence base for Ireland's carbon budgets. The views I express today are my own.

I would like to make five points. First, Ireland's climate obligations are not aspirational targets. They are legally binding limits on cumulative greenhouse gas emissions. Under Irish law, these limits are expressed through carbon budgets and sectoral emissions ceilings. For climate change then, the relevant legal and scientific question is not whether emissions fall at some point in the future nor whether renewable electricity capacity continues to grow. Rather, the question is whether cumulative emissions remain within legally binding limits. Current projections already indicate that Ireland is on track to significantly exceed its carbon budgets and fail to meet a range of climate commitments.

Second, Ireland is among the first countries in the world to experience the consequences of large-scale growth in data centres, which is now accelerating rapidly across the world to serve AI growth. Unfortunately, Ireland is increasingly being viewed internationally as a cautionary tale rather than a model to follow.

Electricity demand growth from data centres has outpaced planning, grid development and the pace at which renewable energy is displacing fossil fuels. The adoption of the Commission for Regulation of Utilities large energy user, LEU, connection policy in late 2025 was an opportunity to align future data centre growth with climate and energy targets. However, my research indicates that it may instead enable significant fossil fuel growth and additional greenhouse gas emissions, placing multiple statutory and policy targets further out of reach.

Third, much of the discussion around data centres and the new LEU policy has focused on renewable energy growth. Renewable investment is positive and necessary, but there is an important distinction between the expansion of renewables and emissions reduction. Growth in renewables only serves the transition to a sustainable and secure energy system if it displaces fossil fuels. Diverting renewables to data centres does not serve this purpose.

A common claim is that data centres support Ireland’s renewable energy transition. The evidence to date does not support that conclusion. Since 2017, growth in electricity demand from data centres has approximately matched or exceeded the growth in renewable generation, meaning that increased renewable supply has not translated into equivalent reductions in fossil fuel use.

Looking ahead to data centre growth under the new LEU connection policy, it should not be assumed that a requirement for data centres to finance renewables projects covering 80% of their demand across a year necessarily means that 80% of fossil fuel generation is displaced without a requirement that demand matches generation in time and that necessary grid infrastructure is in place to deliver renewable energy. Moreover, under the proposed policy, data centres can take six years to reach this condition. During this period, additional demand is likely to be met by fossil gas generation. Therefore, even though the new policy will require data centres to finance renewable projects, it is likely to lead to significant new gas demand and emissions.

Fourth, current national energy and emissions projections do not account for the implications of additional data centre demand that may arise under the LEU connection policy. To help address this gap, I have undertaken analysis of the potential energy and emissions implications of the policy and its potential impact on statutory and policy targets. The scale involved is extremely large. The proposed policy is projected to facilitate 5.8 GW of additional data centre demand by 2040, which is comparable to the scale of Ireland's entire current peak electricity demand. My analysis finds that, if realised, this level of growth would substantially increase electricity demand above that projected by EirGrid, place significant pressure on renewable development requirements, increase fossil fuel consumption during the coming decade and make compliance with carbon budgets even further out of reach.

Finally, I believe the central question is of policy coherence. Ireland has adopted legally binding carbon budgets while simultaneously adopting an economic policy that facilitates a pattern of energy demand growth that is not compatible with achieving them. This is sometimes framed as a debate between economic growth and climate action, a framing I believe is misguided. The issue is whether we are willing to acknowledge that different forms of growth have different consequences and whether those consequences are being honestly accounted for in the development of public policy.

My recommendations are therefore as follows. First, we need to align regulations governing data centres and economic policy more broadly with climate policy. This is why the climate Act applies to all sectors of the economy, including those associated with digital infrastructure and AI. This implies that no new large-scale growth should be facilitated unless it can be demonstrated that it can be compatible with carbon budgets.

Second, we need to accelerate electrification and clean flexibility across the economy. Renewable electricity and electricity infrastructure should be prioritised for displacing fossil fuel use, including transport, heating and industry, while flexibility should be delivered through storage, demand response, interconnection and grid reinforcement. Electrification and clean flexibility reduce emissions and improve energy security. Large-scale, inflexible additional demand, coupled with fossil-fuel generation, does the opposite.

Third, transparency, data collection and independent analysis of data centre energy use, emissions and renewables procurement should be a priority. At present, significant policy decisions are being made without a complete picture of their long-term implications.

I again thank members and I look forward to their questions.

It is good to be here. I thank the committee for inviting me to speak today. I am professor of climate justice at the ICARUS climate research centre at Maynooth University. I am also co-convenor of the Climate Justice Universities Union, an international collective reclaiming the higher education sector as critical infrastructure for just transition and for a climate just future. As an environmental scientist, I have studied interdisciplinary scientific research on climate, energy, environment, technology and policy for over 30 years. Knowing what I know about the ecological devastation of AI, I am conscientious objector. There are people all around the world who are growing in concern and resistance to this.

I will briefly summarise the devastating ecological impacts of AI. The science is very clear. The environmental impacts are dangerous to both ecological health and human well-being. The ever-increasing energy demand of data centres, along with its accelerating carbon emissions, is the most obvious environmental impact. Other impacts include intensive water use that strains water systems, industrial land use that disrupts communities, agriculture and local ecosystems, exacerbated biodiversity loss and a dramatic increase in electronic waste pollution. In addition to that suite of specific environmental concerns, I will emphasise four other, bigger points about the danger of AI.

The first is that AI is preventing the phasing out of fossil fuel. We have known for decades, and the scientific research is clear, that we need to phase out fossil fuels. Unfortunately, the AI hype and the rapid buildout of data centres has led to a dramatic increase in fossil fuel exploration, extraction and use. Fossil fuel industry profits are higher than ever and these large companies have become more effective in convincing governments to expand their subsidies for fossil fuels instead of strengthening environmental regulations to protect human health and the environment. Tech companies are also collaborating with the fossil fuel industry to develop AI tools used to accelerate fossil fuel exploration and allow fossil fuels to be extracted more quickly and cheaply.

My second big point is that AI promotes climate obstruction, which is the denial of climate science, the delay of climate policy and distraction from climate action. I am an internationally recognised scholar who has been studying climate obstruction for over ten years. Our research reveals that large multinational companies strategically and intentionally block climate action to perpetuate fossil fuel use. Tech companies are now deploying many of the same strategies that the fossil fuel industry has used to squash public concern and block climate policy, just as tobacco companies have in the past. To safeguard their profits, these companies lobby against environmental regulations designed to protect ecological health and human well-being. To reinforce their AI hype, the tech companies stifle research, discussion and awareness regarding the environmental impacts of AI.

The third point is that AI is creating a new kind of extractive empire with corrosive supply chains. While previous empires extracted land, water and labour from distant territories to enrich those at the centre, tech companies are now extracting land, water and electricity from countries like Ireland to enrich their senior executives and shareholders. In their quest to build more data centres here in Ireland, tech companies are competing for the same resources that farmers depend on, including land, electricity and fresh water. In his recent encyclical on AI, the Pope reviews how AI is creating vast global networks for extracting resources and how the ecological costs of extraction are being paid by people, communities and households, not by the multinational tech companies.

My final point is that AI is an instrument of dominance, control, surveillance and ecocide. Environmental activists and people who advocate for protecting ecological health are increasingly being silenced through threats and intimidation.. AI, as an instrument of dominance, surveillance and controlling what information people do and do not have access to, contributes to the silencing of those who advocate for more healthy and ecologically healthy futures. This includes women and marginalised people who are experiencing the harms more directly. They are being silenced.

AI is also used extensively in ecologically destructive military applications, including ecocide in Gaza.

In conclusion, there is a growing call for resistance, objection and the banning of any future data centres in Ireland. More and more people are collectively resisting the false narratives of the big tech companies and are objecting to the coercive promotion of AI. If Ireland wants to contribute to a more stable and healthy future, it must stop building data centres and shift our national priorities toward investing in regenerative community wealth-building initiatives and organisations. Continuing to support extractive ecologically devastating companies and their data centres harms public health and endangers Ireland's future.

Thank you very much to all of the witnesses. Before I proceed to questions, I want to acknowledge that today we published our Joint Committee on Artificial Intelligence second interim report. A huge amount of work went into that from all members and it should be available very shortly on oireachtas.ie.

I thank the witnesses. I will start with Professors Stephens and Daly. Professor Stephens identified herself as a conscientious objector and called for resistance, objection and a data centre moratorium. On the question of resistance and objection, there is an argument that the ship has sailed, this stuff is going to happen and even if nation states object or try to resist it, somebody else will do it. What does she say in response to that? Is she essentially saying that we need to stop the train and get off? What does that look like from her perspective?

I thank the Deputy for his question. This speaks to the false narratives that the industry is perpetuating. Many of us in lots of different domains have repeated that we have heard that the train has left the station and, therefore, we need to get on board or will be left behind. The reality is that this is a generated arms race type narrative, to the effect that we have to all get involved, when in fact that is not the case. There is science behind ecological destruction. The way this is going, and the scale and pace, means that things are becoming bigger and faster. People are talking about Manhattan-sized supercomputers. The scale is phenomenal and part of the industry's strategy is to get us to accept the narrative that it is inevitable and we have to keep doing this. If one looks deeper into why this is happening and what problems it is solving and exacerbating, it very quickly becomes clear that it is a false narrative.

I ask the same question of Professor Daly. I ask her to outline her perspective on the idea that we can get off the train at this stage and that is what we should do.

When it comes to AI, the train is just about leaving the station. Ireland's experience with data centre growth is largely not AI driven. It is driven by cloud computing and data centres to serve our financial transactions, storage and the regular operation of tech companies rather than the large scale expansion of AI we are seeing. Ireland does not have to get on the train. On the scale of data centres to serve AI applications across some countries, in particular in the US, I would agree with Professor Stephens that the scale is very difficult to articulate because it is so enormous and is growing so quickly.

Some people have spoken about data centres that can be seen from space.

Yes. For example, the largest data centre I am aware of being constructed at the moment - the growth in this is exponential - is a 10 GW centre, which is much larger than the entire electricity consumption of the city of London. Those can be built within a year and are largely being served with rapidly built and inefficient gas turbines.

The idea that Ireland can compete and should aim to compete in serving this growth is beyond comprehension in terms of our legally binding climate commitments. We can be at the forefront of designing regulation to ensure whatever data centres are built can be compatible. I am not of the view that data centres should never play any role in the future in Ireland. I believe there are policy designs that would enable them to grow in a legitimately zero-carbon way, but at the pace they are currently growing, it is not possible.

Professor Parnell's case is that he does not fall into that bracket and there are benefits of the technological developments in the sector.

The problem is everybody treats AI as one homogenous thing like transport or energy, which is crazy. There are different types of AI. Some, which I mentioned in my report, have a net benefit to society and even reduce emissions compared with what we were doing before. Those should be encouraged and kept separate from the others we are talking about.

There were two points. Professor Daly mentioned one point in 2025 when an opportunity was missed, and an opportunity was missed a couple of years before that when the Commission for Regulation of Utilities was considering a moratorium, which does not mean nothing ever but that you pause while you get ahead of it in regulation and environmental impacts. That was an option considered then but was not taken. There was not and still is not any kind of strict renewable energy requirement attached. Even if you talk about 20% and backup generators, the fact is a net increase in emissions is happening. I am interested in those points and opportunities missed. We are told all the time well do you not like looking at photographs and what about these very good uses, but it is the case that a lot of the AI and data centre uses are in a largely commercial space. I believe Amazon Web Services is the largest user of data centres and developers in the country at the moment. A lot of it is going into making ads appear faster. On Professor Parnell's point about net social benefit, does that need to be part of the regulatory process in terms of expanding, that this is such a dangerous and impactful area of progression that we need a set of tools on environmental impact and net social and environmental benefit attached to each use, and not just at the planning application point? What are the thoughts of the witnesses on that?

It is a really good point. That would be an excellent idea. For example, the CASPIr system I mentioned in my report is a data centre. It is one of 31 funded across the EU as part of this AI factory system. It is used not just for weather forecasting but for protein folding and prediction. Ironically, it is used for lots of our climate change modelling for our predictions of the future. In those circumstances, it has to be said that particular type of data centre is a valid use of our resources and something that actually helps us. There are others which are trickier decisions to make.

That is what I mean. What does getting ahead of it look like? It may involve that kind of moratorium until we get ahead of it.

According to my observations, I agree with Professor Stephens's research findings that many of the narratives supporting the growth of the industry here lie on sustainability claims that are not grounded at all. One claim is that because AI models have been used to reduce emissions in other areas, the energy system being one, the savings found offset the growth in emissions elsewhere, but a recent report, which I can share, looked into each of these findings and showed that the emissions savings associated with AI are associated with a smaller, lighter, tiny segment of the whole AI application, which we should be benefiting from.

AI is a general tool. It is not universally good or bad. In the same way that transportation is a general service, it is not good or bad. Just because an ambulance needs a road to take someone sick to a hospital, it does not mean that we should be building motorways between each city or each town and village. I think that "whataboutism" is quite a shallow argument that is not really evidenced.

I am glad Professor Stephens mentioned water and those other pieces of it. I do not know if transportation is the analogy, because this is effectively a new industry. It was mentioned that, in the next six years, the plan was to actually increase and bed down fossil fuel use, but this is the decade in which we have to achieve our climate action. Regardless of whether a good use case can be made for certain things, the point is that we are adding a new massive industry with profit generation attached to it. How is that balanced out on the remaining carbon emissions space that we have on this planet? How does that balance out when we look to developing countries developing sanitation systems or the many other carbon-related and emission-related uses in the tiny space remaining in terms of carbon? That is the global justice piece. I would ask Professor Stephens to comment on the water piece because I have seen shocking things, particularly from the United States, about entire communities losing water, and we know water scarcity is an escalating issue.

I will come back to the Senator's first question about these narratives that everyone uses it and it is beneficial. It is the exact same playbook that the fossil fuel industry has used. It says that we all use fossil fuels and therefore we cannot phase them out. We know this is deceitful and there is a lot of promoting disinformation and false narratives. There has been an attempt to promote all the research on the social costs of fossil fuels and the social costs of carbon, but they get minimised. The research is there on these other costs, but the industry has captured the policy discourse, so it gets squashed.

In terms of water, these data centres get very hot because they use so much energy and they need a lot of water to cool them down. They are stressing water systems everywhere they are built around the world, including here. The challenge that we have in Ireland in particular is that the monitoring and reporting on how much water they are using is not there. They are very free to just extract water, and they need the water on the hot days, which is the same time when everyone else is being asked to conserve water. The stress and strain on our water systems form a very big issue that is not being adequately regulated and monitored.

I thank all the speakers this morning. Professor Parnell mentioned in his opening comments that we heard an awful lot about the negative aspects of AI in this committee. I was particularly interested when he talked about the weather modelling, his role in that and how Met Éireann had gone from 700 km predicting, to 100 km and down to 1 km. The new forecasting models could be 1,000 times more efficient. Mine was one of the 78,000 households that was affected by Storm Éowyn last year with the inclement weather we experienced on 26 January and we were 13 nights without power. We are still seeing the negative impacts of that all across the west of Ireland in terms of forestry and infrastructure. Anything that can be done to improve the modelling and foresee the dangers has to be looked at in a positive sense.

In terms of weather warnings, we are still looking on county boundaries and how a storm can be red at the Mayo border and all of a sudden it is orange at the Roscommon border, with no barrier there. That is why I was particularly interested to hear Professor Parnell's comments on that. He said that Met Éireann needed more support with the AI models that had a net benefit to society and should not be included in the same category as frontier LLMs.

Will Professor Parnell expand on that a small bit, if he does not mind?

Regarding county boundaries, it is certainly true that weather models, be they traditional or AI, provide more detailed information. The issue there comes from the messaging, and that is something we do not work on; I leave it very much to Met Éireann. How the warnings are conveyed to people is a much trickier political concept and I would not really like to comment too much on it. It is certainly true there is the ability to provide more detailed information. How the public might react to this and what it might mean in terms of emergency services is a much more complicated issue that we would not get into with regard to AI forecasting.

Regarding the differences between AI weather forecasting models and LLMs, most AI systems are made up of neural networks and the mathematics behind them, the data and the computing system. In two of these, the AI weather forecasting models and LLMs are very different. The maths and the neural networks are more or less identical; there are the same things and Senator Scahill might have heard of transformers and graph neural networks. These go into both of them and they are very similar in that respect. Even though the data is taken from the whole world every hour, it is tiny compared with the amount of data going into an LLM. The compute required is a tiny fraction of the amount required to fit something on the scale of an LLM in a data centre that we have just been hearing about. There are similarities but the difference between them is very large.

My next question is for Professor Daly. In a very simple format she pointed out that the increases we all see in renewable energy generation at the moment have not been offset with a reduction in fossil fuels. She made a few points on aligning regulation and accelerating electrification, which we are trying to do. There is also transparency. Breaking it down outside of those three elements, what is her quick ask in terms of how we rectify not reducing fossil fuel consumption?

The formulation of the large energy user connection policy published by the CRU last year requires data centres to be constructed with thermal generation either on site or in proximity to it. This largely means fossil fuels and gas or equivalent battery storage. If we observe the large-scale data centres going through the planning system, they are mainly being built with on-site fossil gas generation plants to meet this condition. The need for the condition arose because of the huge pressure the existing data centre industry has placed on electricity infrastructure. Effectively, data centres now have to build on-site power generation plants so they do not place pressure on infrastructure.

My analysis shows this contradicts various energy and climate policies. To overcome this they would need to build renewables matching 100% of their energy demand from the outset and not six years after energisation. These renewables would have to be genuinely additional and not compete with the renewables we need to decarbonise the rest of our industry, transport and buildings. The regulations would need to make sure the renewables supply and the data centre demand are matched every hour rather than just across the year, so that a data centre does not claim a solar park it finances offsets its gas use in the middle of winter. Finally, the data centre would need to be physically matched with infrastructure to renewables, either through battery storage or grid infrastructure, so the renewables supply could genuinely get to the data centre and it would not rely on on-site capacity because the infrastructure is insufficient. Right now the regulation is designed to meet capacity and flexibility requirements with thermal generation, which is fossil fuel use rather than storage, additional renewables and proper infrastructure.

I thank our witnesses.

Their engagement today is very useful. Professor Parnell outlined an example of the technology not being a goal in itself but, rather, a tool to achieve particular objectives. In terms of the committee making recommendations, supporting small language models or machine learning makes a lot of sense. How does he see the AIMSIR Centre developing? We have been told at this committee, and at the Committee on Climate, Environment and Energy, on which I also sit, that we will see eight severe weather events every year from now on. We cannot predict accurately exactly where they will land. Where does Professor Parnell see his research going? Where does he think we will be with the aid of AI tools within the next two to three years?

Those are great questions. As I said, we have 20 different projects, every single one of which has a student assigned to it and a team of supervisors working with them. I have set them the task of producing one operational deliverable per project. They cover a whole range of topics, not just storm impacts and storm tracking but also looking at floods and rainfall forecasting in the short term. We also have a few projects that are trying to combine weather data with other datasets, including data on trees falling down and power outages, and tying them into medical datasets as well. We know, for example, that climate extremes affect mental health and that floods affect infectious diseases. We are just starting to get to the point where we will be able to tie those datasets together. The great thing about AI is it makes it much easier to tie disparate datasets together and still form predictions. I hope that over the next five years, some - if not most and, I hope, all - of those projects at the AIMSIR Centre will actually give something back and really help people in their daily lives in terms of warnings we give to the public and the way we monitor and predicts things.

We need to be able to show how we can use AI tools to achieve real objectives.

On data centres, there are challenges but, in my view, they are a critical part of our digital infrastructure. All three of the witnesses work for universities that generate large volumes of data, both institutional data and, in particular, research data. Professor Daly might respond to this. Her university, like all the others, stores large volumes of those data in data centres. We have moved away from a situation where universities, for the most part, used storage units on their own campus. We all use those data, for which I assume they are paid electronically. If people watch this meeting back, it will be accessed through a data centre. We use those centres effectively. My challenge to the witnesses is to ask what is the alternative. Do we close down all the data centres that are a critical part of our digital infrastructure and move back to on-site servers in businesses, in homes and on university campuses?

I thank Deputy Byrne for his question. It is really important to clarify that the 5.8 GW of new data centre demand that would fall under the large energy user, LEU, policy is the equivalent electricity generation of six Moneypoint power plants. We do not need six Moneypoint power plants to meet the digital needs of universities, the Oireachtas or the public good. We have spoken about how AI and digital infrastructure certainly are part of our everyday lives. The question is whether unregulated growth of that industry follows logically from the fact we need it. I used the analogy of the transport sector. To get from Cork to Dublin, I use a bicycle, train and electric car. The transport system is essential to all our lives, of course, but does it follow that there should be unregulated growth of motorways or private jets? Where something is a general utility or is called "digital infrastructure", it does not follow that there should be unregulated growth of that industry. History is replete with-----

How do we make those decisions? Development is driven by demand and also by how we use the technology. It is a question of whether it is used for the purposes we have heard about with regard to the AIMSIR Centre and in other areas we are using it, such as in healthcare for predictive purposes. Where do we draw the line in terms of the purposes for which we allow AI to be used?

Before passing that to Professor Stephens, I will say I would question whether the AI applications that are growing energy demand exponentially are genuinely driven by public demand and for the public good. Most of the AI use that has come to me is not because I consciously asked for it.

I want to reinforce the point Professor Parnell mentioned. There are different things being talked about and they are being conflated as part of the industry narrative to push AI. Data-driven AI, including protein folding and cancer research, is very important and we need to support it. What we do not need to support are chatbots and image and video generators - the coercive, exploitative, extractive AI that steals data from the whole Internet and regurgitates it in new forms and controlling ways so that, depending on which AI tool you use, you will get different information.

How as legislators do we draw the line? Where do we draw the line between good AI use and bad AI use?

We have to think about who is being impacted and who is benefiting from this extractive model. It has even been called Epstein tech - some of the stealing by the large language models - because it takes everything from the Internet and it is coercive, exploitative, extractive and does not care about people and what people need. It is being pushed on people. That is where the real concern and resistance are emerging.

I had better keep going but I am sure these issues will recur. The next speaker up is Deputy Murphy.

I will start with a question for either Professor Daly or Professor Stephens. One of the things we have heard from the Government and from AI supporters and advocates is that AI will help us deal with climate change. The KPMG website states, "97% of Irish and global respondents said they believe AI is a net positive for accelerating progress towards net-zero goals". Is that accurate? Do significant positives arise from AI in terms of tackling climate change?

Those claims are largely unevidenced. The research that I have presented and that Professor Stephens has presented outlines material climate impacts we can quantify and demonstrate right now. The theoretical benefits of AI for climate change are largely speculative or unevidenced; if they have occurred, it has been not through this large-scale, large language model but through the smaller and far less energy-intensive models Professor Parnell discussed. So, this is one of those narratives that I think is false.

This future speculation is kind of what the whole AI hype is based on, namely, that once we get this next-level artificial general intelligence, AGI, we can ask it any question and it will solve every problem in the world. It is ideological and not based on science.

It is important to note the research on the climate obstruction of the fossil fuels industry. The way it was effective was through the industry putting itself forward as the solution to the problem, namely, that the fossil fuel industry would create carbon capture and storage and it would solve the climate crisis with its own technology. That is the same playbook big tech companies are using, namely, once they get this advanced, sophisticated, futuristic technology, they will be able to solve these problems. It is a self-delusional reinforcing narrative that is not backed by the science.

The parallels with the fossil fuels industry are striking. I will go back to Professor Daly on the projected impact of the large energy users policy, which is very striking. They could end up using more electricity than the rest of the economy and society combined. Professor Daly put her finger on it in terms of policy coherence. How can it be that the CRU agreed a policy that flies in the face of not just policy, but law in terms of our obligations under the climate Act?

It goes to section 15 of the climate Act, which requires - I paraphrase - State bodies to, in so far as is practicable, align their operations with the achievement of our climate goals. In its finalisation of the LEU connection policy, CRU determined its legal advice was that section 15 did not apply to it.

That is a major determination that will be tested, for example, through the Coolglass judgment. That is a legal interpretation, but we would ask this question: if the CRU is not responsible for operating in a way that meets carbon budgets, then who is?

In Professor Daly's paper, she has estimates about the growth in gas demand, electricity demand and so on. Does she have any figures about what percentage of our carbon emissions data centres will be responsible for if we go on the trajectory that we are currently on? Some 3% to 4% of our carbon emissions are currently coming from data centres.

The table that is appended to my opening statement includes figures on carbon budgets and sectoral emission ceilings. Our electricity sector has a target of reducing emissions to 3 million tonnes of CO2 by 2030 and because we do not have a climate action plan, we do not have ceilings after 2030. For context, according to my analysis, our current annual energy-related emissions are around 35 million tonnes. From my analysis, the additional annual emissions would be 3 million tonnes - that is the entire target of the electricity sector - and 6 million tonnes by 2040.

Senator Higgins spoke earlier about the tiny amount of carbon budget that we have left globally. Ireland, effectively, does not have any carbon budget left in terms of any kind of a fair allocation of the global goal, so, effectively, the scenario my research shows is that this policy will lead to a rise in emissions from the electricity sector rather than a fall, which is contrary to all of the sectoral ceilings and carbon budgets.

I thank the witnesses for coming this morning. My first question is for Professor Daly. There has been a lot of talk about carbon budgets and binding carbon budgets. Who actually sets the budgets? Where does the data come from? Who is involved? What does it look like? How is the data collected, and where is it collected from?

I am happy to go through the process. The process for setting carbon budgets' tracking progress is set out in the Climate Action and Low Carbon Development (Amendment) Act 2021. The Climate Change Advisory Council recommends carbon budgets, according to a series of provisions in the climate Act, including our international and EU commitments. The Dáil then voted on our first two carbon budgets and those were adopted in 2022. The Government then takes the whole economy-wide carbon budget and divides it among sectors through what are called "sectoral emissions ceilings". The Environmental Protection Agency then tracks the progress of what emissions have occurred every year and projects what the emissions will be, according to the policies that have been adopted. Various Ministers are then accountable to the climate action committee to report on progress in their Departments towards meeting the sectoral ceilings. If there is a divergence between the sectoral ceiling and the projected progress, the annual climate action plan is designed to annually update policies and measures to close the gap. The latest climate action plan has been delayed.

My next question is for Professor Stephens. Do we know how many data centres there are in Ireland? How many are under construction? How many have received planning approval?

I do not have the exact numbers. Some of the other witnesses may have them. Does Professor Daly know the number?

The latest count is 100-----

There are currently over 120 plus-----

There are over 100 data centres. This figure combines very small data centres and very large ones.

Dublin has the second largest data centre cluster in Europe. As the witnesses said previously, they are using 21% to 22% of our electricity. They are building more and more, and they are already using 22%. What will their future use be?

It is estimated that it will be up to 30% of our electricity by 2030 or in a few years.

It is causing an increase in all of our electricity bills. The strain on the grid is having a big impact already, and we are not prepared for the scale of projected growth.

To be the devil’s advocate, what does Professor Daly say to the use of nuclear power?

Nuclear power is an interesting proposition. There is no timeline under which we would be able to build nuclear power in Ireland until, probably, the late 2040s, based on the timelines in other countries. By that time, we will have scaled offshore wind and benefited from the resources we already have. We should not plan on the basis of nuclear power because the current technology, or the existing historical technology, is not suitable for our grid. It is too large and inflexible.

There is a new generation of nuclear technology, namely, small modular reactors, SMRs. This is not yet commercially available, however. We do not know what it will cost or whether we can align it with our grid. The Norwegian Government recently commissioned a report to determine whether it would make sense for Norway to invest in nuclear in the same way that is being discussed in Ireland. The conclusion of this process was that it should keep a watchful eye on SMR technology but, in its current state and cost, there was no suggestion that it would meet the energy needs of the country at the moment. The same conclusions apply for Ireland.

Nuclear waste is an issue. Since the issue of nuclear waste remains unaddressed, some people likened building nuclear power to building apartment buildings with no toilets. We do not know what to do with the waste. We do not have a plan in that regard. Therefore, we would not, and should not, do it.

By promoting nuclear energy, we are kind of accepting this rapid growth in energy demand when, in fact, we need to be reducing energy demand if we want to have a stable climate and healthy future. We do not need to perpetuate more electricity and energy. We need to be working intentionally to reduce and move towards sufficiency rather than having this growth mindset where we can continue to grow and grow. The science is clear in this regard. The planetary boundaries are such that continuing to infinitely grow our energy demand, or anything else, is not possible.

It falls to me now to get my five minutes. I will start with Professor Daly. Is it a fair characterisation to say that there is a split in the environmentalist movement between those who see data centres as a draw for further acceleration in renewable investment and those who see data centres as effectively draining the system and not having any linkage to renewable energy systems?

Purely on an observation basis, I see little support in the environmentalist movement - not that it is a single thing - for expanding data centres with fossil fuel use. It is not a-----

I was not really talking about fossil fuel use. I was talking about data centres as a draw for accelerating renewable energy investment. Does Professor Daly agree that there are people, who consider themselves part of the environmentalist movement, who view data centres as a way of accelerating renewable investment?

Again, I do not speak for the environmentalist movement, but the environmental benefits of renewables only accrue if they displace fossil fuels. My research shows that although data centres may be promoting renewable investment, it is not in a way that reduces emissions. Rather, it is in a way that grows energy and diverts renewable energy from displacing fossil fuels. There may be an overlap between, let us say, the renewables industry and the environmental movement in this regard. The growth of data centres is good for the renewables industry because it needs customers, but it is not necessarily good for environmental objectives. Does that make sense?

Yes. There are a lot of people sitting at home watching this who dip in and out of their understanding on this. Everyone understands the climate emergency, and I think a lot of people understand it more than we think. Perhaps that is often unfairly reflected in the media. For example, for a lot of people in this country, the Green Party is considered to be at the vanguard of the environmentalist movement.

Its leader, Eamon Ryan, when he was the Minister for the environment, publicly stated that he recognised that investment in data centres was a catalyst for further renewable investment. I do not know what his position is now or what his position is on the large energy-user action plan, LEAP, and I am not trying to ascribe a position to him, but there is a big difference between what he said and what Professor Daly and Professor Stephens are saying.

In his time as Minister, Mr. Ryan directed Gas Networks Ireland to stop offering new connections to data centres to facilitate on-site gas generation. As far as I am aware, as leader of the Green Party, he was clear that building new gas generation capacity to serve large-scale data centres was not aligned with climate policy.

Five offshore wind farms are under planning. We hope two will be approved this year with three to be approved next year. Those offshore wind farms will provide a total of 6 GW. It is one thing getting planning approved, as we know from housing in this country, but it is quite another to get them up and running and connected to the grid. Does Professor Daly think those offshore wind farms could be viable in the absence of large energy user customers such as data centres?

Yes. This was why I emphasised in my opening statement a recommendation that the Government create customers for those offshore wind farms by having a very strong electrification policy. That would include the electrification of transport, including road freight, of existing industry and of buildings. At the moment, the plan is that a portion of the customers of those offshore wind farms will be data centres. That assumes we will not go fast enough in electrifying other sectors of the economy, which means that those sectors will still be reliant on fossil fuels and we will, consequently, have high bills. The solution is the electrification of the rest of industry. My own research through the energy modelling work I referenced shows that we need multiple gigawatts, perhaps 10 GW or 15 GW, to meet our climate goals. We have the potential for 50 GW or 60 GW. When there is a surplus of offshore wind, it could certainly serve new industries, such as large energy users, but if that capacity is in the short term diverted to new demand customers rather than electrifying our existing fossil fuel applications, it will mean failure to meet our climate targets.

I will turn to Professor Stephens. I am trying not to abuse my position as the Vice-Chairperson but I could see she wanted to say something. Before she does, I will ask if her position is different to that of Professor Daly in the context of the generation of new electricity demand. Her last bit of testimony was that we should be trying to reduce electricity demand. That seems different to what Professor Daly just said.

Our positions are aligned. My point is that if we want our energy transition to move away from fossil fuels to a renewable-based future, we will use electricity and energy differently. The research I have just been involved in, funded by the Sustainable Energy Authority of Ireland, SEAI, was about community-based renewable energy generation. This is where Ireland has missed the boat in focusing primarily on these large-scale commercial renewables while not investing enough in community-owned and household-level renewable generation that is local and appropriately mixed for communities. We definitely do not need AI and data centres to create demand for renewables. There is plenty of demand for energy around the country. We need to ramp up investment in regenerative renewable electricity generation throughout the country. That is separate from data centres and AI. We do need them to create the demand. The demand is there for what people need.

I apologise as I had to step out, but I did hear the witnesses' statements at the outset. If my questions are repetitive, they should please feel free to answer briefly. I will look back at the proceedings of this session if what I wish to raise has been covered already. Professors Stephens and Daly both made really strong points in their opening statements about the greenwashing that is going on, particularly around data centres. I have a couple of questions on this specifically. One of the lines we hear about data centres is that they can be run with renewable energy. There is a big push from the companies to have generation by private wires, and they cite that as a green solution. Does that stand up to scrutiny when they represent such a large amount of energy consumption?

Energy costs in Ireland are so closely tied to oil and gas prices, and we have seen how other countries, like Spain, have taken action to move away from that. What impact are the large energy users connection policy and its use by AI data centres having on us moving away from fossil fuels and on our household energy bills? I ask the witnesses to respond for a minute each on those questions, if that is okay. I have one other question, for Professor Parnell.

As was mentioned, Ireland is now held up internationally as a cautionary tale about the strain on the grid and the increase in electricity or energy prices for every household around the country because of data centres. We have to pivot in terms of the strategy because this is just going to keep getting worse unless we do so. That is really important to acknowledge.

The new legislation, as I understand it, states 80% of the new data centres in six years, or sometime in the future, will have to use renewable energy. This does not make sense because 110% should be renewable. The companies should actually be helping us to move toward renewable energy, not doing a little bit but then still exploding and increasing fossil fuel demand in Ireland. That is the opposite of what is required. We have been a leader in fossil fuel phase-out internationally in some domains, but when it comes to what is actually happening on the ground, we have not invested in fossil fuel phase-out. That is actually what we need to be doing, and that requires a pivot in the whole strategy on data centres.

I will address the private wires question and what constitutes responsible data centre development. There is a narrative that data centres that are not placing pressure on the electricity grid represent what is responsible development, but if development comes with on-site gas power plants, that runs contrary to our energy security and climate objectives. That has not been accounted for in the public discourse or in policy development, as far as I am aware, which is a major gap. The narrative does not stand up to scrutiny.

There was a recent report on a data centre in the Dublin area. It was reported internationally as the first data centre in Europe, possibly the world, to be built with a microgrid, meaning no connection to the grid at all. This was lauded as responsible, but it is being built with its own power plant. It is not straining the grid, but it is shifting the pressure to the gas network and to our carbon budgets.

On the impact on bills and energy prices, I have not done research. Friends of the Earth published a report on this last week that contains the first quantification. One of the reasons Ireland's electricity bills are so high is that gas sets the price a lot of the time, because we are still very dependent on it. We have the highest electricity prices in Europe and our dependence on gas is one of the main reasons. If we continue to build gas-fired power plants to serve marginal electricity generation, that trend will continue.

I thank the witnesses.

I have a final question, for Professor Parnell. In his opening statement, he mentioned the open-source nature of the technology used for weather predictions. Has he any concerns that there is a danger regarding this technology because it is in the hands of such a small group of people in the US. In the US, as we know, the National Oceanic and Atmospheric Administration, NOAA, has essentially been captured, and in some areas in the US people have to pay to access extreme weather information. Is that something Professor Parnell has concerns about because of the private nature of who holds this technology?

Maybe in other countries, but I cannot see it in Europe. We have the amazing European Centre for Medium-Range Weather Forecasts, which is funded by 23 nations and provides public forecasts up to two weeks ahead, which we can take, tailor and deliver to the public. I do not believe we are in danger of what the Deputy described. There is a related-----

Sorry, but many people in the US would have said they did not think they were in danger of it either, but that is, unfortunately, the route we have gone down. I appreciate that the Trump Administration has been a major factor in this regard, but unfortunately the issue is that the technology, so long as it sits in the hands of private companies, has the potential to become a commercial asset.

The witnesses have described that as life-saving information. It is difficult.

AIFS, for example, is open source. Therefore, we can see how much it costs to train it and can make decisions about its benefits. We do not have such information from many other large language models. We do not know how much electricity or water they are using, which makes it very hard for all of us to make any kind of responsible decision about them.

I thank Professor Parnell.

I do not normally sit on this committee and appreciate being accommodated today. I sit on the climate committee. Reference was made to the Government's industrial policy and how it is linked with data centres. A KPMG report commissioned by the Government was released last week and stated that data centres and, presumably, to an extent, AI are linked to job creation and the anchoring of industry in this country. Do the witnesses have a view on that? Does any research reference the number of jobs genuinely or directly linked to data centres?

The short answer is that it is not an area my research covers, but the same report had estimates of emissions associated with the growth of data centres in line with projections. The assumptions were somewhat different from what I have in this report. The report showed a substantial number of growing emissions with data centre use. They do not demonstrate how that can be aligned with legally binding commitments. I do not have information on jobs.

There have been many headlines about huge job losses because of AI. I am concerned that the Government is getting bad information. I read the KPMG report. This is part of the industry strategy. What has happened regarding research is that big tech has moved from how research is done. Rather than peer-reviewed research, there is increasingly PR-reviewed research. It is about the research that promotes the hype. That is what is being fed to Governments, including the Irish Government. I am concerned that this erodes trust in our democracy, universities and the research being done in universities. Some of it has the same narrative about promoting AI hype and squashing any research that is critical. This is really dangerous.

I am Irish but lived in the United States for a long time. I moved back to Ireland in part to make sure the things that I saw happening with research and academia in the United States did not happen here, where there is a very strong public system of research and education. It is very important that universities do not become captured by industry interests which are trying to promote growth in their technologies. This involves chat bots and generative AI, not data driven AI.

Are the type of AI or systems used by data centres monitored or regulated anywhere in the world? In practical terms, it is something that we, as legislators, will struggle with. Are there any practical ways of monitoring and regulating these things?

To my knowledge, there are no regulations governing what data centres are doing. As part of the public discourse in Norway about the potential impact of data centres on Norwegian bills, an idea was floated about knowing what purpose data centres were serving. If the industry is promoting a narrative that it is growing on the basis of public good or utility, it should follow that it is transparent about what it is doing because there are good and bad uses of AI.

In the academic HPC area there is a clear route. You have to apply, fill in a form to say what you are using the HPC for, how much time you are using and what you are doing. There is a very clear, open way of seeing if it is of value or not.

The industry has strategically invested to block regulation and accountability. It speaks to its power. It is growing. The rapid scale of AI and data centres, and the devastation and destruction they are causing around the world, speaks to the lack of regulation and accountability. It is a real problem.

It was mentioned earlier there is no regulation or monitoring of the water use of data centres in Ireland. Is that correct?

That is what I understand. They can use as much-----

We talk a lot about their energy usage and the other issues with them but we do not do that with water.

I thank Deputy Ahern. I was going to raise at the end of the meeting that we might write to Irish Water to clarify that, with the agreement of the committee.

And to the CRU about whether it has a regulatory role also.

We will write to those two entities with the agreement of the committee. We will do a second round of three minutes. First up is Senator Higgins.

I want to follow up on a couple of points. We touched on it but I want to come back to the piece about budgets. It is not a matter of whether it is good, although it is notable that Veritas and Friends of the Earth research estimates that 52% of data stored in data centres is dark data, and we do not know what it is, and over 30% is obsolete or trivial. That is about 82% that is not delivering in any shape or form. On the bigger question of the global environmental piece, I sit on the climate committee so I see the very limited space we have to act. I would like to get a sense of the time urgency not just in the abstract and the general but the importance of caution now, specifically over this five- to ten-year period when we are at a crunch point or tipping point in terms of climate and the remaining uses in that regard.

On the water piece, we know that some of the very large users and drivers of AI do not care about climate change as a fundamental risk. Elon Musk and the former Google CEO, Mr. Schmidt, have said we should go all in because we are never going to meet our climate goals anyway. Peter Thiel has said that "People are spending way too much time thinking about climate change, way too little thinking about AI." There is a narrative that is quite hostile to climate action coming from some of the leaders in the AI industry. A large amount of the resources are centred in the United States, which has left the Paris Agreement.

Will the witnesses comment on the interconnection piece? What are we connecting into? What is its footprint? The stories from the United States are very extreme in terms of data centres. It is not unconnected. Another point of connection the witnesses might comment on is LNG. I know that Professor Daly has written about it. Data centres are being cited as one of the rationales for the importation of LNG in planning applications in Ireland. What is the danger of us embedding or facilitating systemic pieces? That comes to the piece around the new colonialism frame. I think that is something to be aware of. Will the witnesses comment on those points?

I thank the Senator for the opportunity to articulate how precarious, putting it very mildly, the current climate situation is and how this is helping to push us in absolutely the wrong direction. The Paris Agreement binds countries to doing what they can to limit global warming to 1.5°C because the impacts are so incredibly devastating. We are talking about a shutting off of the AMOC - the north Atlantic stream - which would change Ireland's climate utterly. That is just the impacts on Ireland. The impacts felt all around the world are unfathomably bad. We have already effectively crossed the 1.5°C threshold, which means any emissions we put into the atmosphere now will have to be drawn back down at some point in time.

The carbon budgets are what link Ireland's specific policies with that global agreement. On the basis of our current trajectory, we are already far off track to meet our carbon budgets. My research shows that growing data centres in the way that is being planned will exacerbate that problem. Professor Stephens's research shows how the indirect applications of AI are helping to spread obstructionism to climate as well, but my research focuses on how it puts Ireland's commitments out of reach.

I know we are due to hear from the CRU. When I posed a specific question to it on this matter, it said there is no evidence to directly suggest that data centres in Ireland are contributing to higher electricity prices. It said that there is a considerable ask on the grid but because of the financial contributions data centres make we cannot directly attribute the two. It is important that we deal with some of the facts around these issues. I am of the view that data centres are a critical part of our digital infrastructure. I totally agree with the comment that was made earlier that there is a difference between what we might call good AI and bad AI. We need to make decisions on what we are using this new technology to power.

I want to talk about the relationship between our digital infrastructure and the drive for renewables, because I think they are going to be linked. I am looking at what China is doing at the moment. It has already rolled out a wind-powered underwater data centre that uses far less energy as an environmental solution to some of this. It has just announced a $295 billion investment in its data centre infrastructure. Part of that is around securing its own digital infrastructure.

By the way, I totally agree with Professor Stephens's concerns around AI being used for surveillance purposes, particularly in terms of how China will do it. China's approach is quite interesting. It is very clearly marrying its investment in digital infrastructure with a major investment in renewable energies and particularly in innovative ways of moving in that space. One of my big frustrations here is how long it is taking us to meet our renewable targets, particularly on the offshore side. The point was made about the train leaving the station. There is a global train leaving the station. If Ireland separates its need to invest in digital infrastructure from its need to invest in renewables, we are going to be in trouble and will be left behind. Perhaps one of the witnesses can address these points, particularly about China.

I thank Deputy Byrne for the question. I also saw the article on the underwater data centre. I am delighted that we can produce things like that. I think it is amazing. However, I worry - as Professor Daly or Professor Stephens mentioned - that when something like that is built, it takes away from another opportunity for electrification. The money could have been spent on something which actually reduced the overall emissions, rather than being spent on something else which is an extra data centre. I have a slight concern about those kinds of areas, but in general if we can do things like that, it is wonderful.

I thank everyone for their really interesting contributions. I am delighted that we are getting to start this discussion on climate and AI. I want to talk about the AI Act, because Professor Parnell referenced it in his contribution. He explained how the tools he is talking about are not covered. We also heard recently in our defence module that defence is not covered either by the AI Act. People have this kind of separation in their heads of what AI is versus what online social media platforms are, even though we know that these platforms and their algorithms are powered by AI. They are regulated separately. I would argue that they have not been effectively regulated. We are now seeing an AI Act which was potentially ambitious but is now being gutted by the digital omnibus and does not go as far as Chinese law and regulation in a lot of ways around consumer protection and so on. It also does not cover climate to my knowledge or understanding of the legislation.

What does regulation look like that captures all of these in a more comprehensive way and deals with the fact that AI, as we have heard, is not homogenous and comes in all shapes and sizes? I ask Professor Stephens to comment on this.

The problem here is that we have this extreme concentration of wealth and power with these big tech companies. I do not think this technology and this industry is like any other technology or any other industry ever before. The companies have captured Governments to push it, repeating the narratives they have very carefully articulated, including that the train has left the station, the inevitability narrative and the narrative that if you are against it, you are against progress and that it is absolutely essential. In terms of climate obstruction delay, they are very intentionally making it a race with urgency, which creates chaos.

What does better regulation look like? I have my issues with the AI Act but what is the perspective of the witnesses as climate experts?

I will take this question. Ireland's example is very instructive. It is that if data centre demand grows faster and outpaces renewable growth, then it will grow with fossil fuels. This works against our climate policy and energy security commitments. It is a simple matter, although maybe not simple, of policy alignment. It is about aligning the EU AI Act and Ireland's data centre expansion targets, and aligning all economic policy on this with our climate obligations.

There is a board for the AI Act at EU level. My understanding is that it is comprised of representatives from departments of enterprise and finance from across the EU, so it does not have capacity on fundamental rights protections, for example. Does Professor Daly know whether it discusses issues such as climate?

It is beyond my expertise and I do not know.

I reiterate what Professor Daly mentioned, which is that we do have climate legislation and laws and we have human rights laws and legislation. We have been wilfully ignoring them and saying we cannot use these laws with regard to AI. Upholding the legislation we already have is fundamental.

My first question is for Professor Daly. A key thing she said in her opening statement, which is something the proponents of more and more data centres and more and more AI like to ignore, is that "[s]ince 2017 growth in electricity demand from data centres has approximately matched or exceeded the growth in renewable generation". Would Professor Daly say the growth in data centres is a predominant reason we are not coming close to meeting our climate targets? We add more renewable energy, and perhaps it is insufficient, but we keep expanding our usage.

Currently, I would say "No". The EPA has published a report to state we are just about within our first carbon budget and the compliance issue will be in the future. The trend Deputy Murphy raised, which is that demand from data centres outpaced renewables generation, contributed to more emissions but not to the same extent as the failure to provide public transport or the growth in agricultural emissions. Other sectors are much larger. When I undertook the research that came up with the finding, I was not that concerned in the big scale of things about data centres in the context of Ireland's carbon budgets. However, it is now, when 5.8 GW is projected to grow potentially for six years, initially on the basis of fossil fuels, that it will create a significant compliance gap. It is the projection forward that will create the biggest compliance concern rather than the history.

It gets worse and worse. My next question is for Professor Stephens. I like the conscientious objector framing. It is a way of asserting that we have choice because they are attempting to eliminate any choice. AI is not neutral technology; it is a technology which has the values of venture capitalism, which are deeply anti-humanist, embedded in it. What can we do? The most simple thing we could do policy-wise is to have no more data centres in Ireland. I think that is what we should do and it is really clear.

However, even if that is done, there still could be a pushing of AI onto people across the country. We could have no more data centres here but we would still be expanding our energy usage, affecting clime globally and having a negative impact. What other measures can we take? One would be a right for people to opt out of AI. We are forced into using AI in various ways in terms of our work, accessing public services and just using the Internet. It is kind of pushed on us. Does Professor Stephens have any other proposals as to how we can get off that train?

I see two ways in which the Government has been promoting AI. One is by supporting the data centres and the other is by pushing it in all our public services and in education and higher education. The AI Ready programme, from what I have read about it, encourages people to sign up to generative AI tools such as Gemini and ChatGPT. It is actually promoting those chatbots, which are the bad AI. They are the ones that are bad on so many levels. We need to create spaces within society to acknowledge the harms and that we all have agency and do not have to use AI if we do not want to do so. In particular, encouraging older people to use those tools is causing all kinds of angst and frustration. The Government has been promoting that and Microsoft has been integrally involved in developing the Government programme to get everybody on board.

It is important to mention the gender dimension that once again arises. It is widely acknowledged that women are more sceptical and less willing to use these tools. Yet, I have heard big tech executives say that getting more women to use more AI tools is exactly what the AI Ready programme should do. Governments should not be promoting technology like this in that way.

I hope Professor Stephens finds our general scepticism to be an example of being gender non-conforming, but perhaps not. I am marking off all the names here.

I have some questions on Professor Daly's responses to Deputy Murphy. If I understood correctly, her position is that the growth of data centres in Ireland to date has not necessarily been one of the key contributors to our not meeting the targets we want to meet, with the sectors having a bigger impact being agriculture and transport. Her concern is with the potential trajectory of data centre growth in Ireland in alignment with the LEU policy. She mentioned growth in demand to 5.6 GW but that is a market estimate of demand and does not necessarily mean there will be construction in Ireland of 5.6 GW of data centre capability. How does she square that with the policy that 80% of generation must be reliant on renewables? Going back to the point I was trying to tease out earlier, is it Professor Daly's understanding that it is technically or economically viable to construct a data centre that relies 100% on renewables? My understanding, and she can correct me if I am wrong, is that some of it could be done with software or there could be a reliance on hydrogen and all those kinds of things but the technology is not really there yet at scale.

The changes we are making on climate in this country benefit everyone living in Ireland. There is no doubt in my mind about that. Ultimately, however, we are part of a global project in which the world is acting in solidarity to reduce global warming. Climate sceptics often make the argument that Ireland is only a small part of that and why should we care when all the impacts are being created by China, etc. That leads to complacency and is, in any event, fundamentally wrong. It is wrong because action in Ireland benefits everyone living here and it is fundamentally wrong from a humanist perspective, which Deputy Murphy mentioned in the context of AI. Can the same be said when it comes to data centres? If those centres are going to be constructed, is it not better that it be done in jurisdictions like Ireland that are promoting renewable usage rather than, for example, allowing those who wish to construct them to abuse poorer countries or, indeed, countries that do not have progressive policies on renewable energy? I am interested to hear Professor Daly's observations on those points.

I thank the Leas-Chathaoirleach for giving me an opportunity to discuss what I think is the main misconception in regard to the LEU connection policy and the 80% renewables requirement.

That requirement only kicks in six years after the data centre is constructed. In the meantime, it is likely the data centre's marginal generation will be fossil fuels. The point of our climate legislation is not more renewables at some time in the future; it is minimising cumulative greenhouse gas emissions as quickly as possible to keep within a budget. We are already over budget.

I shared with the committee a working paper in which I model the impact of the 5.8 GW under that six-year glide-path period in which the data centre runs on fossil gas. It will thereafter meet 80% of annual demand with renewables. The findings are stark. It shows that electricity-----

Sorry to interrupt, but is there a chicken-and-egg situation here? If we say the data centre must have 80% renewable right now, it has to build the renewable thing first, presumably, and then charge up the data centre. Is the point not to get the benefits of renewable investment aligned with the potential benefits of having data centres and a digital infrastructure in Ireland?

If you build the data centre before you build the renewables, then you increase gas use and greenhouse gas emissions. The scale that is in the paper is very substantial. It means a growth in gas demand for electricity generation, rather than a decline in line with current policies.

Have you assessed the comparable growth in renewable energy?

Yes.

What does that look like?

It looks as though there will be no growth in renewable energy by 2030 because of the glide-path period. After that, it will require 20 TW hours of renewable electricity by 2040, which is greater than the current renewable electricity generation. It would grow renewable electricity but in a way that does not serve our climate goals because-----

It would take too long. Is that the point?

No. Our renewables targets meet our climate goals because they are designed to displace fossil fuel use now, not just to serve demand growth. A big concern is that the projections by the EPA that inform development of the climate action plan, which is supposed to take corrective measures should our targets be off-track, do not include the impact of the LEU connection policy. This paper is the first time, to my knowledge, these impacts have properly been quantified. It shows a big lack of alignment between policies.

Professor Daly said Ireland had adopted legally binding carbon budgets but also said it had adopted an economic policy which basically means using more fossil fuels. Obviously, these conflict with each other. How can we achieve a fair balance between the two?

The Climate Change Advisory Council has a statutory obligation to advise the Government on policies in its annual review. It has consistently in its annual review of the electricity sector advised against what has been adopted, which is the construction of fossil fuels to serve growth. The council's advice is that new data centres should be facilitated only when 100% of their demand is matched by additional renewables from the outset, such that we are not growing fossil demand early and then using renewables after a glide-path period.

Professor Parnell differentiated between AI weather models using large language models and the ones he uses based on data sets. In what way is there less energy in that process? Why is it not included in the AI law?

The data sets are far smaller than AI weather forecasting models. You are talking gigabytes versus - I could not even quantify how much goes into a modern LLM like ChatGPT-5. Under the AI Act, these are not deemed critical infrastructure unless they start to control warnings or other systems. In terms of weather forecasting, whether AI or physics based, there would always be a human layer before the warning is processed and given out to different people. You would never have a system that automatically generated and fed out warnings.

There would always be expert meteorologists at Met Éireann who would say, "Well, let's see about this before we give anything to the public". It would only be in that case that it would trigger the AI.

I am just not sure. Believe it or not, there is a developer from my town, Enniscorthy, who developed a national flood prediction app. It is free in part, but it also charges for extra features. Deputy Gibney brought up a situation that she says did not happen, but it is actually happening. What is Professor Parnell's opinion on that?

Personally, I am delighted to see things like that. The weather data that we collect is of such high quality that anybody who has a good bit of skill in terms of data science can go and try something like this, but the wonderful thing about weather forecasting is that it is benchmarked every single day. I find out tomorrow whether my weather forecast was any good or not, and I can constantly keep checking, so if people come up and develop something that is better than what we are doing, then I am happy to adopt that. What that person needs to do is adopt a thorough benchmarking strategy, which, when I checked yesterday, still was not on the website. However, I would be delighted if he produced something better.

It is brilliant, because it came out of the catastrophe of the town being flooded for weeks. He was told that it could not be developed or would take ten years to develop it, and he developed it in a couple of months.

Again, that needs to be checked and benchmarked before we can say it is a really useful product. I certainly would not pay for it yet. I am still going to trust Met Éireann and its expert guidance.

Senator Alice-Mary Higgins has promised me that she is going to take less than 15 seconds.

In terms of private wires and the issue of energy security, the witnesses mentioned that there may be renewable energy and it is feeding into data centres and large energy users, but that it may not actually be connected to the grid or even have an obligation to share with the grid in extreme weather events, for example. Will the witnesses comment on that piece?

Will they also comment on the information tools piece?

Ah here, come on, Senator.

I mentioned the issue of the owners, but there is also the issue related to gamification. There is huge money going into gamifying the results of ChatGPT, etc. In terms of climate denial, is there an issue with the very tools that the AI is supporting?

If they do not mind, will the witnesses just answer the first question, given that Senator Higgins is getting in another round of questions?

Will the Senator restate the question?

It is about the private wires legislation and-----

If the witnesses do not want to answer that, I can just go to the second question.

I will just say that it is really important that we keep infrastructure public and do not have a separate grid for corporate interests. Keeping public infrastructure public is a fundamental part of democracy. Extreme heat, drought, floods and fires are coming, and the CEOs of the big tech companies all have private bunkers that they have built. They will all be okay. We have to think about people who are already marginalised and vulnerable, and this is where the injustices and inequities are getting exacerbated by these approaches.

On the cheery note of Armageddon bunkers, we will conclude this meeting. I thank all the witnesses. This was a very constructive meeting and it added to the discourse, which is what this forum is about. I thank everyone for their participation.