A Board and CEO Briefing on Climate Risk, Artificial Intelligence, and the Case for Immediate Action
Executive Summary: Climate Risk Has Become a Systems Intelligence Problem
Environmental warming is often framed as an energy transition problem — a matter of emissions targets, renewable capacity, and climate finance. All of these forces matter, but they obscure an equally important structural issue: the institutions responsible for anticipating and managing a warming planet are still relying on tools built for a climate that no longer exists.
The data make the challenge impossible to dismiss. The Copernicus Climate Change Service confirms that 2025 was the third-warmest year on record globally, at 1.47°C above the pre-industrial baseline. More significant than any single year, the three-year period from 2023 to 2025 was the first in recorded history to average more than 1.5°C above pre-industrial levels — the threshold the Paris Agreement was designed to hold. “The world is rapidly approaching the long-term temperature limit set by the Paris Agreement,” Copernicus director Carlo Buontempo has said. “We are bound to pass it; the choice we now have is how to best manage it.”
The Global Carbon Budget 2025 report puts the trajectory in stark terms: fossil-fuel CO2 emissions reached a record 38.1 billion tonnes in 2025, atmospheric concentrations are approaching 426 parts per million — 52 percent above pre-industrial levels — and the remaining carbon budget consistent with a 50 percent chance of holding warming to 1.5°C now amounts to roughly four years of emissions at the current rate. “With CO2 emissions still increasing,” the report’s lead author Pierre Friedlingstein has concluded, “keeping global warming below 1.5°C is no longer plausible.”
The policy response has not closed the gap. The United Nations Environment Programme’s Emissions Gap Report 2025 finds that even full implementation of every country’s current climate pledge would still leave the world on a path to roughly 2.3 to 2.5°C of warming this century; current policies alone point toward as much as 2.8°C. On the current trajectory, the report concludes, the multi-decadal average is likely to exceed 1.5°C by the early 2030s.
The economic translation of that physical trend is already visible on corporate and government balance sheets. Swiss Re Institute reports that insured losses from natural catastrophes reached USD 107 billion in 2025 — the sixth consecutive year above the USD 100 billion threshold — driven not by a single mega-disaster but by the relentless accumulation of severe convective storms, floods and wildfires that insurers call “secondary perils.” Los Angeles’s wildfires alone set an all-time record for that peril.
The arithmetic is unforgiving. Boards cannot legislate their way to a lower global temperature, cannot insure their way out of a risk that is growing faster than premiums can reprice it, and cannot manage what their existing systems cannot see quickly enough to act on. Something in the intelligence infrastructure that sits between climate science and corporate and government decision-making has to change.
This is where artificial intelligence — applied to forecasting, emissions monitoring, materials science and early warning — moves from an interesting research capability to a strategic necessity.
AI models such as Google DeepMind’s GraphCast now outperform the world’s leading physics-based weather forecasting system on the large majority of variables tested, at a fraction of the computing cost. Climate TRACE, an AI and satellite-driven coalition, now independently tracks more than 660 million individual emissions sources worldwide, covering roughly 75 percent of global emissions without waiting on self-reported government data. FireSat, an AI-powered satellite constellation backed by Google Research, can now detect a wildfire the size of a classroom before it becomes a headline. None of these systems slows the physics of a warming planet. All of them close the gap between what is happening to the planet and what institutions can see, predict and act on in time to matter.
That gap-closing role — not a substitute for emissions reduction, but the fastest available lever for managing the risk while the harder work of decarbonization continues — could make artificial intelligence one of the most consequential tools available to boards and governments navigating the next decade of climate risk.
1. The Climate System Is Outrunning Our Institutional Tools
For decades, climate risk management has rested on an assumption that is now demonstrably false: that the recent past is a reasonable guide to the near future. Insurance pricing, infrastructure design codes, agricultural planning and corporate risk registers were built on historical records of temperature, rainfall and disaster frequency. That approach functioned reasonably well in a stable climate. It is failing in a non-stationary one.
That is where the current model begins to strain.
The World Economic Forum’s Global Risks Report 2026 captures the shift starkly. Over a two-year horizon, geopolitical and informational risks dominate executive attention — geoeconomic confrontation, disinformation, societal polarization. But over a ten-year horizon, the picture inverts: extreme weather ranks as the single most severe long-term risk facing the world, and environmental risks — extreme weather, biodiversity loss, critical changes to Earth systems, natural resource shortages — occupy five of the top ten positions. Roughly 57 percent of respondents expect the next decade to be “turbulent” or “stormy.” The risk that keeps compounding is precisely the one shortest-term incentives are least equipped to price.
Swiss Re’s 2025 loss data shows what that compounding looks like in practice. Severe convective storms alone caused an estimated USD 50 billion in insured losses — the third-costliest year on record for that peril and part of a clear multi-year upward trend. Wildfires caused roughly USD 40 billion, with the Los Angeles fires accounting for the majority of that total and setting a new record. The United States absorbed 83 percent of global insured losses in 2025. “Insured losses keep rising,” Swiss Re’s group chief economist Jérôme Jean Haegeli has noted. “Strengthening prevention, protection and preparedness is essential to protect lives and property.”
It helps to think about how the tools available to see and manage this risk have actually evolved, rather than treating “climate technology” as a single category. The earliest stage was pure observation — weather stations, tide gauges, satellite imagery — recording what happened without predicting what would happen next. The second stage was physics-based modeling: supercomputers running numerical simulations of the atmosphere and ocean, powerful but slow and enormously expensive to run at high resolution. The third stage, now well underway, is predictive AI: machine-learning weather and climate models trained on decades of physical simulation data, delivering forecasts in minutes rather than hours at a fraction of the computational cost. The fourth stage, still emerging, is ambient, continuous monitoring — AI systems that do not wait for a scheduled forecast run but instead watch emissions, hazards and ecosystems in near-real time and surface risk the moment it appears. The furthest stage, and the one boards should be planning toward rather than assuming, is anticipatory action: early warning systems wired directly into grid dispatch, evacuation protocols, insurance triggers and supply-chain rerouting, so that AI-detected risk produces an automatic, governed response rather than another report nobody reads in time.
Most institutions, if they are candid about where they actually sit on that curve, are still working from stage one or two data inside stage-three risk models — historical loss tables and physics-based projections — while the tools now available operate at stage three and four. That gap between the risk intelligence an organization could have, and the risk intelligence it actually uses is where the costliest surprises tend to originate.
CINDY’S PERSPECTIVE
We need to stop treating climate risk as purely an environmental issue owned by a sustainability team. It is increasingly a systems-intelligence problem, and it belongs on the same enterprise risk register as cyber risk and geopolitical exposure — reviewed with the same rigor, and resourced accordingly.
If the physical climate is now moving faster than our historical-record-based models can track, leadership’s task is to close that gap with the best available intelligence tools rather than wait for the next disaster to reveal how far behind the organization already is. That means funding AI-driven monitoring and forecasting as core risk infrastructure, not as a sustainability pilot project competing for scraps of budget.
2. Climate TRACE and GraphCast Are Showing What AI-Powered Climate Intelligence Looks Like in Practice
Two initiatives illustrate, in very different ways, what AI-powered climate intelligence looks like once it moves out of the research lab and into continuous global operation: Climate TRACE and Google DeepMind’s GraphCast and WeatherNext family of models.
Climate TRACE began in 2019 as a modest Google.org-funded effort by WattTime and TransitionZero to track power-plant emissions from satellite data. At former U.S. Vice President Al Gore’s suggestion, it grew into a coalition of more than 100 nonprofits, universities, technology companies and research institutions. Its premise is deceptively simple and quietly radical: rather than waiting on national governments to self-report emissions — often years late, and frequently estimated rather than measured — Climate TRACE uses artificial intelligence to analyze more than 90 trillion bytes of data drawn from over 300 satellites and 11,000 sensors, independently tracking more than 660 million individual emissions sources worldwide and covering an estimated 75 percent of global emissions at the level of individual facilities and assets, refreshed monthly.
That is a fundamentally different proposition than the emissions inventories most boards currently rely on. It replaces a system built on national self-interest and multi-year reporting lag with one built on independently verifiable, near-real-time observation — the same shift toward radical transparency that auditable, explainable AI has driven in other high-stakes domains.
GraphCast and its successor WeatherNext models take a parallel approach to physical forecasting. Trained on decades of historical weather data using graph neural networks rather than traditional physics simulation, GraphCast has been shown to outperform the European Centre for Medium-Range Weather Forecasts’ benchmark system on more than 90 percent of tested variables and lead times — and on 99.7 percent of variables in the troposphere, the atmospheric layer where most weather that matters to people actually happens. It generates a ten-day global forecast in under a minute on a single processor, work that previously required hours on a supercomputer running hundreds of machines. In practical terms, that means earlier and more accurate warning of cyclone tracks, the atmospheric rivers that drive catastrophic flooding, and the onset of extreme heat — the difference, in a given event, between an evacuation order issued with days of notice and one issued with hours.
Neither system changes the physics of a warming planet. Both dramatically shrink the time between a risk emerging and an institution being able to see it clearly enough to act.
CINDY’S PERSPECTIVE
What interests me most about Climate TRACE isn’t the satellite technology. It’s the decision to make the entire emissions inventory public and independently verifiable, rather than proprietary.
Climate accountability is trust-sensitive for good reason: capital allocation, regulatory compliance and public credibility all depend on numbers people can actually check. Building AI systems that are transparent and auditable by design takes more discipline than building ones that are simply accurate — and it is exactly the same discipline responsible AI in any high-stakes domain requires.
Boards evaluating climate-intelligence vendors should be asking the same question about every one of them: can this system’s outputs actually be audited, or are we simply being asked to trust a dashboard?
3. AI Is What Turns Climate Data Into Preemptive Action
A forecast attracts attention because it is visible and immediate. The more consequential shift is what AI is beginning to do upstream and downstream of the forecast — discovering the materials the energy transition depends on, and detecting the hazards that outrun human observation entirely.
On the detection side, FireSat — an AI-powered satellite constellation developed by Google Research with the nonprofit Earth Fire Alliance, satellite manufacturer Muon Space and launch partner SpaceX — is designed to close one of the most consequential gaps in wildfire response: the minutes and hours between ignition and detection. Its sensors, combined with AI models trained to distinguish genuine fire signatures from noise, can identify a fire as small as five meters by five meters, catching low-intensity ignitions that existing satellites miss entirely. A pilot satellite launched in 2025; three more followed in July 2026, moving the system from proof of concept toward continuous global coverage — funded in part by more than USD 15 million from Google.org.
On the materials side, Google DeepMind’s GNoME system illustrates how AI can accelerate the physical inputs the energy transition depends on, not just the information layer around it. In a matter of days, GNoME identified 2.2 million candidate inorganic crystal structures — roughly a tenfold increase over the total number of stable inorganic materials known to science after decades of human research — and more than 700 of those candidates have since been experimentally validated, with an autonomous robotic laboratory achieving a 71 percent synthesis success rate on a test set. Among the candidates: 25 times more potential solid-state lithium-ion conductors than previously known, and new compound classes that could reduce dependence on cobalt in battery cathodes. Materials discovery has historically been one of the slowest steps in bringing better batteries, solar cells and energy-storage technology to market; AI is now compressing years of that search into weeks.
On the grid side, one of the earliest and best-documented examples remains one of the most instructive: Google’s DeepMind team applied machine learning to 36-hour-ahead wind-farm output forecasts and increased the value of that wind power by roughly 20 percent, simply by making its output predictable enough to commit to the grid confidently in advance. Predictability, as much as raw generation capacity, is what determines how much intermittent renewable energy a grid can actually absorb — and that is a forecasting problem AI is well suited to solve at scale.
The emerging pattern across all three examples is the same. AI does not generate energy, and it does not extinguish fires. It compresses the time between a physical signal and an actionable decision — early enough for a grid operator, a battery manufacturer or a fire crew to actually do something about it.
CINDY’S PERSPECTIVE
This is the point at which CEOs need to think beyond funding another pilot. A wildfire-detection model is a research project; a wildfire-detection model wired into dispatch protocols, insurance triggers, and public alert systems is operational risk infrastructure — a fundamentally different commitment, and a fundamentally different level of accountability.
Too many organizations fund an interesting AI proof of concept and call it climate action. Real impact happens when operating procedures, capital planning, and emergency response actually change around what the model can now see.
Boards should be asking a bigger question than “what pilots are we running”: which of our physical assets — plants, supply chains, real estate, insured portfolios — would benefit today from the kind of early-warning intelligence that already exists, and why hasn’t it been deployed yet?
4. Climate AI Is Moving From Pilots to Planetary-Scale Infrastructure — With a Real Cost
Climate TRACE and GraphCast show what governed, transparent AI-driven climate intelligence looks like in live operation. FireSat and GNoME show autonomous detection and discovery moving from research into infrastructure. The broader picture, however, requires boards to hold two facts in tension at once, because AI’s relationship to the climate crisis is not simply that of a rescuer.
The International Energy Agency’s 2026 analysis of energy and AI is unambiguous on the scale of the challenge. Global data center electricity consumption reached 485 terawatt-hours in 2025, with AI-focused facilities growing 50 percent in that year alone; the IEA projects total data center consumption will roughly double to 950 terawatt-hours by 2030, reaching about 3 percent of global electricity demand, with AI-focused facilities tripling their own consumption over the same period. Some advanced AI workloads — video generation, complex multi-step reasoning — consume hundreds or even thousands of times more energy per query than simple text generation, a trend that can offset hardware efficiency gains faster than they are achieved.
That is the uncomfortable half of the picture. The other half is that the IEA’s own analysis finds well-documented AI applications in industrial optimization and grid management alone could save more than 13 exajoules of energy by 2035 — equivalent to roughly 3 percent of current global final energy consumption. Note: an exajoule is a scientific unit of energy and an exajoule is equal to one quintillion joules (10¹⁸ joules)
AI, in other words, is simultaneously a meaningful and growing source of emissions and one of the more promising tools available for reducing them elsewhere in the economy. Both statements are true, and a board briefing that presents only one of them is not a complete briefing.
This tension is now showing up explicitly in how the world ranks its own risks. The World Economic Forum’s Global Risks Report 2026 places extreme weather as the single most severe risk facing the world over the next decade — and “adverse outcomes of AI” in the top five of that same list. The technology increasingly seen as essential to managing the physical risk is, in the same breath, seen as a systemic risk in its own right.
That is likely how this actually resolves. No organization gets to treat AI-for-climate as a straightforward net positive without also accounting for the electricity, water and hardware footprint of the AI systems it deploys. The path forward runs through AI applications sited on genuinely low-carbon power, procured and measured with the same rigor applied to any other major capital or emissions decision — not through treating AI’s climate benefits and AI’s climate costs as separate line items owned by different teams.
CINDY’S PERSPECTIVE
The question is no longer whether AI can help detect, forecast, and respond to climate risk faster than the tools we had a decade ago.
It can. The more important question is whether the organizations deploying it are honestly accounting for their own energy footprint in the same breath as its benefits — or quietly booking the upside while leaving the electricity bill for someone else’s carbon inventory.
This is why leaders should resist climate-AI theatre. A model that performs impressively in a vendor demo is not climate action. A monitoring system that measurably shortens warning time, reduces losses, and is transparent about its own footprint, audited the same way its outputs are audited, is climate action.
The category needs far more accounting and far less announcement.
5. Governance and Trust Could Become the Real Differentiator
The governance opportunity may ultimately be as consequential as the technical one.
The Emissions Gap Report 2025 contains a warning with direct relevance for any board relying on AI-generated climate intelligence: only 60 parties to the Paris Agreement, representing 63 percent of global emissions, had submitted updated climate pledges by the report’s September 2025 cutoff. Institutional accountability for the underlying climate trajectory is already stretched thin. Layering AI-generated forecasts, risk scores and emissions claims on top of that thin accountability, without clear standards for how those outputs are validated, does not close the credibility gap. It risks widening it.
An AI model that helps a utility forecast wind output is relatively low-risk if it is wrong. An AI model that underpins a company’s public emissions claims, a municipality’s evacuation threshold, or an insurer’s decision to withdraw coverage from an entire region carries a different order of consequence entirely — reputational, financial, and, in the case of evacuation and insurance decisions, potentially life-safety consequences. Few organizations have explicitly governed for that distinction, because the tools arrived faster than the policies did.
This is where explainability stops being a research nicety and becomes the actual product of a credible climate-AI strategy. A model whose reasoning and data lineage can be inspected can be defended to a regulator, an auditor, an insurer, or a community relying on its warnings. A model that simply outputs a number, however impressive its marketing claims, cannot — and “the model said so” will satisfy neither a board nor a courtroom when a warning arrives too late, or an emissions claim turns out to be wrong.
The human dimension deserves equal weight. As AI absorbs more of the detection, forecasting, and monitoring workload, the scarce and valuable capability that remains is judgment: the willingness to act on an early warning before certainty arrives, the authority to reallocate capital toward resilience before a loss forces the issue, and the accountability to stand behind a model’s output rather than hide behind it. Those are leadership qualities, not technical ones, and no amount of forecasting accuracy substitutes for them.
If governance and transparency are designed into climate-AI systems from the outset, rather than added after a missed warning or a disputed emissions claim, the old trade-off between speed and trust could begin to weaken — institutions could see further ahead and remain more accountable for what they do with that sight, not less.
CINDY’S PERSPECTIVE
For too long, climate governance and AI governance have been treated as two separate conversations, run by two separate teams, reporting up two separate committees.
The greater opportunity is to fuse them: explainable models, auditable data lineage, and clear human accountability for every consequential decision an AI system informs — designed in from the start, not bolted on after the first public failure.
That is a far more durable proposition for boards, regulators, insurers and the communities depending on these systems to get the warning right — and it is a genuine competitive advantage for the organizations that build it first.
Executive Conclusion: Climate Action Is Entering Its Intelligence-Augmented Moment
Artificial intelligence will not lower global temperatures on its own. Emissions reduction, energy transition, adaptation finance and international cooperation all remain the primary levers — but none of those levers removes the immediate requirement to see this risk clearly enough, and early enough, to manage it.
The physical trajectory is no longer seriously contested. 2025 was the third-warmest year on record and part of the first three-year period ever to average above 1.5°C. The remaining carbon budget for a 50 percent chance of holding that line is now measured in a handful of years, not decades. Insured catastrophe losses have exceeded USD 100 billion for six consecutive years. That creates a simple operating reality for every board with physical assets, insured exposure or a supply chain: the risk is compounding faster than most current intelligence systems can track it.
Climate TRACE offers one of the clearest examples of what independently verifiable, AI-driven emissions transparency can look like at global scale. GraphCast and WeatherNext are showing how AI forecasting can outperform physics-based models on speed and accuracy simultaneously. FireSat and GNoME demonstrate that the same approach extends from early warning through to the materials science the energy transition depends on. The International Energy Agency’s own data is the necessary counterweight: none of this comes free, and AI’s own energy footprint has to be managed with the same discipline applied to the risks it helps detect.
Collectively, these signals suggest that climate risk management is beginning to cross an important threshold — from static, historical, backward-looking models toward continuous, AI-driven, forward-looking systems capable of closing the gap between a warming planet and the institutions responsible for responding to it.
For CEOs and boards, four directions follow directly from the evidence in this briefing. First, treat climate risk intelligence as core enterprise infrastructure, reviewed with the same rigor as cyber risk, and fund AI-driven forecasting and monitoring accordingly rather than as a sustainability-team side project. Second, before adopting any AI-for-climate tool or vendor claim, demand the same explainability and auditability this briefing has argued for throughout: if a model’s reasoning cannot be inspected, its output should not be the sole basis for a material decision. Third, account for AI’s own energy and water footprint inside the organization’s climate liability, not as a separate line item managed by a different team — the electricity bill for the intelligence layer is part of the cost of the intelligence. Fourth, move deliberately from pilots to production: the technologies described in this briefing are not laboratory curiosities; they are operating today at global scale, and the gap between what is technically available and what most organizations have actually deployed is now the more urgent risk to close.
The strategic question for boards is therefore not whether artificial intelligence has a role to play in managing environmental warming. The evidence in this briefing says it already does, at a scale most institutions have not yet absorbed. The question is whether today’s leaders will build the governed, transparent version of that capability before the next record year forces the issue.
Cindy’s Closing Perspective
I have spent much of my career arguing that the value of artificial intelligence lies less in its raw capability than in how honestly and transparently that capability is deployed. Climate risk is the domain where that argument carries the highest stakes I have encountered.
The pattern is familiar by now: first the technology looks unrealistic, then it becomes interesting, then the scale of what it reveals becomes genuinely uncomfortable, and eventually it becomes part of how serious institutions operate. AI-driven climate intelligence is moving through those stages quickly, because the physical evidence is no longer leaving anyone the luxury of a slower timeline.
We should not overstate what technology can do. A forecasting model cannot replace the hard political work of cutting emissions, cannot make a poorly governed institution act on a warning it receives in time, and cannot substitute good judgment for an organization unwilling to invest in resilience before a loss forces its hand.
But complacency here would be the far greater mistake.
The remaining carbon budget for 1.5°C is now measured in years. Insured losses have cleared USD 100 billion for six straight years. The tools to see this risk coming, clearly and early, already exist and are operating today at planetary scale. These realities create a powerful reason to treat climate intelligence with the same seriousness boards already apply to financial and cyber risk — not less.
The organizations that deserve the closest attention are therefore not simply those with the most impressive climate models. They are the ones willing to make those models transparent, govern them honestly, and act on what they show before the next record year decides for them.
Perhaps the defining question for leaders is no longer whether they can afford to invest in AI-driven climate intelligence. Perhaps it is whether they can afford to keep managing a warming planet with tools built for a climate that no longer exists.
Notation: In my newest book, co-authored with Cathy Cobey, Global Head of EY AI Assurance, The AI Precipice: An Executive Guide to Balancing Innovation and Safety, we discuss the crucial moonshots to achieve a better world. Advancing deeper international strategies to combat climate change is no longer an option – it’s a necessity. Every company has a voice and needs to lead with more courage.
Lady Whistledown’s Climate Society Papers
Dearest Gentle Reader,
It has come to this author’s attention that the planet has, once again, run rather a high fever, and that the physicians called in to examine it are, for the first time in recorded history, machines.
How very modern.
The patient’s chart, dear reader, makes for sobering reading. Three consecutive years averaging above the very limit the world’s leaders solemnly promised, in Paris, never to cross. A carbon budget so nearly spent that the finest minds in the field now describe it, without any apparent enjoyment, as “virtually exhausted.” Insurers — not known for sentimentality — have now paid out more than one hundred billion dollars in disaster claims for six years running, and Los Angeles alone contributed a record sum to that unhappy total.
Enter the algorithms.
One consortium, egged on by a former vice president with a famous slideshow, has quietly built a machine that watches the entire planet’s smokestacks so governments no longer have to be asked politely to confess their own emissions. Another has taught a computer to out-forecast the finest supercomputers in Europe using, by this author’s understanding, rather less equipment than a well-appointed kitchen. A third has launched satellites clever enough to spot a fire the size of a classroom from outer space, which seems the sort of thing that would have gotten someone burned at the stake in a less enlightened century.
How scandalous. How entirely necessary.
There is, naturally, a delicious complication. The very data centers powering all this planetary vigilance are, this author is reliably informed, becoming rather significant consumers of electricity themselves — so significant that the World Economic Forum has ranked “adverse outcomes of AI” as a top-five threat to civilization in the very same report that ranks extreme weather as the worst threat of all. One rather suspects the algorithms and the atmosphere are now competing for the same electrons.
There will naturally be those who declare the machines have arrived just in time to save us, and those who declare they have arrived just in time to finish the job. This author suspects the more interesting truth, as usual, lies in who is left holding the electricity bill and the accountability, once the applause for the demonstration has died down.
The greater disruption may therefore not be the arrival of the forecasting model itself.
It may be the realization that a warming world can, in fact, be watched — provided somebody, somewhere, remains willing to act on what the watching reveals, and to answer for it honestly when the model’s carbon footprint comes due alongside its benefits.
One should watch the transparent players particularly closely. Society is endlessly enthusiastic about a dashboard once someone else has done the auditing. It is rather less generous to the institutions that must first prove, in public, that the number on the screen can actually be trusted.
Leadership, dear reader, always appears most obvious in retrospect.
For now, the satellites are watching the smokestacks, the models are outrunning the supercomputers, and some of the world’s most consequential decisions are quietly being informed by machines that, one hopes, someone is still checking.
The thermometer, as they say, is still rising.
Until our next gathering, dear reader,
Lady Whistledown
Research Bibliography
Best Practice AI. “DeepMind Increases Value of Wind Power by 20% by Predicting Supply 36 Hours in Advance.” Case study.
Climate TRACE. “About the Coalition.” climatetrace.org.
Cobey, C. Gordon, C. The AI Precipice: An Executive Guide to Balancing AI Innovation and Safety, Routledge, 2026.
Copernicus Climate Change Service (C3S). “Copernicus: 2025 Was the Third Hottest Year on Record.” 2026.
Global Carbon Project / Future Earth. “Global Carbon Budget: Fossil Fuel Emissions Set New Record in 2025.” 2025.
Google DeepMind. “GraphCast: AI Model for Faster and More Accurate Global Weather Forecasting.” Corporate blog.
Google Research. “Three New FireSat Satellites Launch to Help Detect Wildfires with AI.” Google blog, 2026.
International Energy Agency (IEA). “Key Questions on Energy and AI.” World Energy Outlook Special Report, 2026.
Swiss Re Institute. “2025 Marks Sixth Year Insured Natural Catastrophe Losses Exceed USD 100 Billion.” Press release, 2026.
United Nations Environment Programme (UNEP). “Emissions Gap Report 2025.” 2025.
VentureBeat. “Google DeepMind’s Materials AI Has Already Discovered 2.2 Million New Crystals.” 2023.
World Economic Forum (WEF). “Global Risks Report 2026: Top Risks in an ‘Age of Disorder.’” 2026.
World Meteorological Organization (WMO). “State of the Global Climate 2025.” WMO-No. 1391.
#AI #AIBoardGovernance #AIClimateChange #AIEnergy #AIGovernance #AIStrategy #AIClimateRisk #BoardGovernance #CarbonFootprint #ClimateChange #ClimateRisk #ClimateTRACE #CopernicusClimateChangeService #GlobalRisks #EnergyFootprint #GNoME #Google #GraphCast #LadyWhistledown #ResponsibleAI #TrustedAI #UnitedNations #WeatherNext #WindPower #WorldEconomicForum #WEF #WorldMeteorologicalOrganization #WMO
