AI safety fears mount as ex employees warn of extinction risk

Jacob Coxon, a former Anthropic and OpenAI researcher, posted that the industry is racing toward self improving superintelligence and gambling with our lives⁷. His resignation was echoed by Anthropic safety lead Evan Hubinger who warned there could be over a 10 percent chance of human extinction within the next decade¹¹. These warnings spread as Coxon and colleagues urged regulators to slow development and demanded better alignment, transparency, and governance across AI labs^7,13.

The controversy intensified as Coxon’s messages went viral, prompting public discussions about how far labs will go to push new models, and whether safety work can keep pace with speed of progress^17,29. Officials and activists weighed calls for international limits, with policy figures seeking closed briefings and new rules to slow what experts say is an out of control race².

The debate also touched on broader costs, including data center energy use and environmental damage, a point raised by researchers linking AI growth to climate and resource pressures³. The conversations show a fault line between market incentives to deploy powerful tools and the need to protect people from unpredictable AI behavior as the technology edges toward general and even superintelligent forms^9,21.

In the coming weeks, more resignations, policy proposals, and potentially new safety frameworks are expected as the industry and governments try to answer who should set the pace and how to keep humans in control^12,18. Coxon and other signatories have called for measurable safety milestones, independent verification, and transparent risk assessments to accompany rapid lab progress, while labs argue that fast iteration is essential for maintaining competitive advantage and for delivering practical benefits to society, a tension that underpins ongoing regulatory scrutiny, industry unity, and public trust as researchers push toward ambitious capabilities^7,13,17.