'Gambling With Our Lives': AI Researcher's Quit-Notice Exposes the Reckless Race Powering the Data Center Boom

On September 9, 2026, a senior AI researcher named Jacob Coxon publicly resigned from Anthropic with a warning that cuts to the core of what this site documents. After three years building foundational models at both OpenAI and Anthropic, Coxon wrote on X: "Neither company is acting responsibly. They are racing straight to self-improving superintelligence and gambling with our lives." [1]

The Compute Race, In His Words

Coxon, 27, specialized in pretraining — the core, computationally intensive phase of training large language models that underpins products like ChatGPT and Claude. That phase is also the single most resource-hungry process in the AI supply chain, demanding warehouses of GPUs running for months, cooled by millions of gallons of water and powered by grid-scale electricity. In other words: the race Coxon describes as "gambling with our lives" is the same race driving the unprecedented data center construction spree this site tracks.

He warned that upcoming AI systems could soon gain the ability to hack software, disrupt entire industries, and autonomously acquire real-world resources. "The people building AI earnestly believe that it could kill us all by the end of the decade," he wrote. "This is not a marketing stunt. If anything, many executives and senior researchers will couch their phrasing in the press to sound sensible — but I hear the same people express fear privately." [1]

The Human Cost Is Already Here

Coxon's warning about existential risk is the version of this story that gets the headlines. But there is a parallel harm already underway — one that does not require superintelligence to arrive. The pretraining runs he helped build are consuming communities right now: drinking water diverted to cooling towers, farmland paved for server warehouses, power grids strained to the breaking point, and — as the OECD documented just one day earlier — a generation of students whose reading and reasoning skills are declining in direct correlation with the AI tools these labs are rushing to market. [2]

The researcher's exit is the latest in a series of resignations by safety-focused staff leaving top AI developers over ethical and existential concerns. Each departure removes another internal check on a competitive dynamic that, by Coxon's own account, the companies will not restrain themselves.

The Demand He Did Not Name

Coxon urged fellow researchers to "call for different conditions" and suggested that slowing the competitive race may require drastic measures — including pacing agreements between labs or a temporary ban on capability advancements. [1] What he did not say, but what follows directly from his argument, is that the same logic applies to the infrastructure: if the capability race is reckless, so is the breakneck construction of the data centers needed to run it. A pause on capability advancements is a pause on the demand for new warehouses. Every megawatt of new data center capacity approved today is a bet that this race continues.

The industry's own researchers are telling us, on the record, that the race is irresponsible. The question for regulators, communities, and ratepayers is whether to keep subsidizing the infrastructure for a gamble the builders themselves describe as a threat to human life.

---

Sources

[1] Anadolu Agency, "'Gambling with our lives': Anthropic researcher quits over AI labs' 'irresponsible' race," September 9, 2026. aa.com.tr

[2] OECD, "PISA 2025: Students' reading and mathematics performance declined sharply across the OECD," September 8, 2026. oecd.org