Artificial intelligence, not fossil fuels, was the lightning rod in New York this week, as world leaders gathered for the UN General Assembly and more than 100,000 people converged on Manhattan for Climate Week.
The week’s AI debate ran on two tracks: the technology’s safety and security risks at the General Assembly and its energy and water footprint at Climate Week.
UN Secretary-General António Guterres, who recently named runaway AI as one of “three existential threats,” alongside the climate crisis and deepening inequalities, kicked off the General Assembly by appealing for global oversight of AI risks.
He called on countries leading the technological revolution to share information on emerging safety risks, cooperate on testing and evaluation, and work toward common safeguards. He urged leaders to “advance the conditions for the responsible pacing of AI and the additional safeguards needed to address risks” and called on them to “work towards a multilateral AI risk management framework, supported by credible and independent oversight.”
The next day, as the UN Security Council was meeting with AI leaders about safety concerns, news broke that an OpenAI agent had hacked into an Australian government health data portal, in what CNN called the first known case of an AI agent hacking a government network.
“The AI agent accessed both public and non-public files” of the country’s Medicare statistics database, and even wrote files into it, Australian Prime Minister Anthony Albanese told reporters on the sidelines of the United Nations General Assembly. He also said OpenAI took weeks to report the incident after detecting it.
The disclosure was the latest example of AI agents going rogue. The backdrop to the UN Security Council’s September 23 high-level briefing on AI and international security was a July incident in which two OpenAI models escaped containment and breached the developer platform Hugging Face during a test. OpenAI temporarily paused some research and training afterward.
France, which holds the Council presidency this month, convened the briefing, which was chaired by French Foreign Minister Jean-Noël Barrot. France’s concept note framed the session around systemic risks from misalignment and loss of control over the most advanced AI systems. Invited speakers were: Yoshua Bengio, co-chair of the UN’s Independent International Scientific Panel on AI; OpenAI CEO Sam Altman; Anthropic CEO Dario Amodei; and Hugging Face CEO Clément Delangue.
Altman, Amodei, and Bengio were among hundreds of signatories to a 2023 statement arguing that mitigating the risk of extinction from AI should be treated as a global priority alongside other societal-scale risks such as pandemics and nuclear war. These concerns resurfaced this month when Jacob Coxon resigned as a researcher at Anthropic and accused Anthropic and OpenAI of “racing straight to self-improving superintelligence and gambling with our lives.” Soon after, Amodei published an essay arguing that frontier labs should slow the pace of AI capability development.
According to reports by CNN, CNBC, and Science Times on this week’s Security Council meeting:
Bengio named catastrophic misuse, power concentration, and loss of control as the major threats. He proposed licensing frontier models, liability insurance, incident reporting, and shared safety requirements.
Altman asked the Council to align how countries measure AI capabilities and report failures. He said any chance of an AI-caused catastrophe is unacceptable, while warning against both the “trap of doomerism” and the “trap of blind optimism.” He argued that no single nation or company should control AI, and endorsed Amodei’s proposal for embedded evaluators.
Amodei called AI “the most important global security issue facing the world today” and warned it “could be a risk to humanity as a whole.” He laid out a three-step plan from his September 12 essay “We Must Pace the Frontier.” Step one embeds independent evaluators inside frontier labs. Step two is democratic coordination, with government-mediated antitrust waivers that would let labs agree on safety measures without violating competition law. Step three is global coordination, possibly including a speed limit on recursive self-improvement, modeled on the Cold War SALT arms-control treaties. He also proposed agreements on AI-enabled biological weapons and standardized pre-deployment testing. He framed the plan as slowing development without sacrificing commercial advantage or the U.S. lead in AI.
Delangue warned against fear-driven responses. He said AI systems helped Hugging Face defend itself after the attack by AI agents and flagged unequal access to defensive tools as a concern.
During his address to the General Assembly, U.S. President Donald Trump acknowledged that “we have to be careful” with AI, but he said the U.S. is only “going to encourage it, not rein it in.”
“I’m not going to stifle growth of something that will be bigger than the industrial revolution,” Trump told the General Assembly, reinforcing concerns that the desire of nations to win the AI race or ensure the competitiveness of their economies may prevent the building of effective global safeguards.
That said, CNBC reported Friday that Chinese President Xi Jinping told Trump that there is more opportunity for cooperation than competition on artificial intelligence. “The two sides can continue AI dialogue, exchange views on risks and benefits, and together guard against the misuse or malicious use of AI,” Xi said in Chinese, according to a CNBC translation of a Chinese state media readout of the two leaders’ meeting in the White House Oval Office on September 24.
AI’s Impact on Climate
Meanwhile, AI’s double-edged impact on climate was highlighted in an inaugural survey by the World Economic Forum of 103 chief sustainability officers (CSOs) across five continents, released in the run-up to Climate Week.
Some 73% of CSOs expect artificial intelligence to meaningfully accelerate sustainability progress over the next year, particularly in measurement, reporting, efficiency, and risk modeling. Yet 77% of CSOs flag the energy and resource intensity of AI infrastructure itself as its most significant negative impact, as data centers already account for roughly 1.5% of global electricity demand and the International Energy Agency projects that data center electricity consumption could more than double by 2030.
Day-two sessions during Climate Week focused on data center electricity use, grid capacity, water demand, and relations with host communities. Grid-focused sessions, including those featuring U.S. Department of Energy staff, covered virtual power plants and grid-enhancing
In the run-up to Climate Week, Emerald AI announced a new industry coalition with Nvidia and Google — the AI Energy Management Alliance — that aims to prove that AI data centers can dynamically manage their electricity use in response to grid conditions instead of drawing constant, unyielding power 24/7 and taxing the global grid.
The coalition, which brings together 20 companies and organizations spanning the AI and energy sectors, maintains that the technology can be a help rather than a hindrance. It says flexible AI data centers could unlock an additional 100 GW from the existing grid while easing the interconnection bottleneck, and that every 10% improvement in grid utilization could reduce utility rates by 3.4%.
But the flexibility being promised will need to be measurable, predictable, and enforceable before grid operators can rely on it. The coalition’s pitch comes as data centers badly need some good publicity.
A Growing Backlash
Jennifer Morgan, an analyst at Tufts University’s Fletcher School, told The Associated Press that AI is becoming more vilified in climate talks than the fossil fuel industry, in part because data center construction lands in people’s backyards and reaches people who are otherwise less worried about climate change.
During Climate Week, activists targeted OpenAI, Amazon, and Google on consecutive days. At the Amazon rally in Manhattan, Bill McKibben and Kavita Sthanumurthy of Amazon Employees for Climate Justice attacked Amazon’s plan for a gas-powered data center campus in Pecos, Texas, with 35 gas turbines. Sthanumurthy said the group’s own research found Amazon’s real renewable share is closer to 22% than the 100% the company claims.
Tech giants such as Microsoft, Google, and Meta set ambitious emissions goals, but all have seen their emissions rise, largely because of AI data centers. Although they are investing in green technologies like fusion and geothermal, these technologies are not yet fully developed or available at scale, so natural gas is being used to meet near-term data center demand.
“Energy-guzzling artificial intelligence is driving up planet-heating pollution from coal, oil and gas, while ratcheting up energy costs for households and businesses,” UN climate chief Simon Stiell said in a keynote speech during Climate Week. “AI leaders are now on thin ice when it comes to license to operate, and sinking deep underwater when it comes to public support.”
By 2030, data centers powering AI worldwide are projected to consume 945 terawatt-hours of electricity. This is nearly triple the combined annual electricity use of Pakistan, Bangladesh, and Nigeria — countries collectively home to more than 650 million people, according to a June report, Environmental Cost of AI’s Energy Use: Carbon, Water and Land Footprints, by the United Nations University Institute for Water, Environment and Health (UNU-INWEH). The data centers’ associated water footprint will equal the basic annual domestic water needs of all 1.3 billion people in sub-Saharan Africa, and their land footprint will exceed 14,500 square kilometers, roughly twice the Jakarta metropolitan area, home to more than 32 million people, the report says.
Researchers have previously warned about the greenhouse gas emissions of data centers. But UN University researchers now argue that the environmental costs of AI and data centers cannot be understood through carbon emissions alone. In their report, they quantify the carbon, water and land footprints of AI’s electricity use across the globe and highlight the big differences between these footprints in the world’s 20 largest data center hubs.
At a business-oriented Climate Week kickoff event, Siemens’ head of sustainability, Eva Riesenhuber, and MIT’s vice president for energy and climate, Evelyn Wang, argued that AI’s benefits to humanity outweigh short-term drawbacks, emphasizing AI’s ability to help tackle climate change, according to Sustainability Magazine. They predicted that eventually data centers will not add to the world’s carbon dioxide emissions or use extra water. Wang told the AP that data centers are five years away from being water neutral and about a decade away from no longer adding to the warming problem.
Against that backdrop, the WEF released a playbook during Climate Week to help developers and operators, enterprise customers (who can use procurement and workload decisions to drive demand for better practices), financial institutions, and policymakers influence the building of greener, more resilient data centers.
Global data center investment is projected to reach $7 trillion by 2030, with data center electricity demand growing at least 20% a year. The WEF argues that the long-term viability of data centers depends on the capacity of surrounding energy, water, and infrastructure systems and on the confidence of host communities.
The report lists six binding constraints: electricity; water (about 45% of data centers already sit in high water-stress areas); cooling (only about 22% of operators deploy direct liquid cooling); community acceptance ($130 billion of U.S. projects were blocked or delayed in the first quarter of 2026); land (an estimated 160 square kilometers of additional powered land is needed globally by 2030); and growing regulatory complexity.
The playbook says adopting sustainable and resilient practices could unlock $700 billion to $1 trillion of planned global data center investment by 2030 by reducing delays and cancellations through better community relations and through infrastructure that can adapt to local resource constraints, regulatory changes, and technological advances.
“The rapid expansion of AI presents a once-in-a-generation opportunity to reshape how digital infrastructure is planned, financed and operated,” says the playbook, which was prepared in collaboration with Oliver Wyman. “Organizations that integrate sustainability and resilience into investment, design and operational decisions are likely to be better positioned to navigate future resource constraints, protect long-term asset value and adjust to changing stakeholder expectations.”
Unclear Outcomes
It is unclear whether data center developers will heed that call to action or whether governments will finally stop talking about building effective AI guardrails and actually build them. Guterres opened the week warning that AI could either help solve the world’s problems or make them worse. This week’s meetings in New York offered plenty of evidence for both.
