Former Anthropic and OpenAI Researcher Resigns, Accusing Companies of Knowing A.I. “Could Kill Us All by the End of the Decade” – Clarín | 09/09/26

Jacob Coxon, a 27-year-old British researcher who spent the past three years working on artificial intelligence pretraining at OpenAI and later Anthropic, resigned from the industry and accused both companies of irresponsibly racing toward self-improving superintelligence despite being aware of its potential risks. Coxon argued that people developing these systems genuinely believe A.I. could kill all humans by the end of the decade and warned that no company should develop technology exceeding human capabilities without government intervention or a coordinated slowdown. His concerns were supported by Anthropic employees: safety lead Evan Hubinger personally estimated a greater than 10% risk that A.I. could kill all humans within the next decade and said there is still no plan to solve the alignment problem for superintelligence, while Samuel Marks said A.I. developers believe the technology could cause human extinction or similarly severe outcomes. A central concern is “recursive self-improvement,” a scenario in which a model could design its own successor beyond human control. In February, Anthropic removed a commitment from its safety policy to halt model development if risks could not be controlled, while in late July more than a thousand industry employees — including Anthropic CEO Dario Amodei — called on Washington to support technical and governance mechanisms that could deliberately slow A.I. development in response to emerging risks. Coxon argues that companies are trapped in a competitive race to develop these systems first and urged researchers to challenge that dynamic and demand different conditions, warning that moving toward superintelligence without adequately understanding how it works represents an excessively dangerous gamble. Link to Article