Recommended
Joe Benton, who worked in artificial intelligence safety research at Anthropic, revealed that he has left the company and issued a warning about the risks he believes are posed by the race to develop increasingly intelligent systems. His departure comes just days after the resignation of researcher Jacob Coxon, who also expressed concerns about the direction of AI.
READ ALSO: AI could kill all humans before 2030, warns OpenAI and Anthropic engineer after resigning
Benton announced on his X account that he left the safety team at Anthropic, the company behind Claude, approximately two weeks ago and that he will now work outside of the major artificial intelligence labs to promote independent evaluations and greater transparency.
His message contains a particularly blunt warning. Benton maintains that companies are competing to develop machines "much smarter than any human" and believes there is a possibility of catastrophic consequences.
"We may not survive this," he wrote.
Who is Joe Benton and what did he do at Anthropic?
Benton was not an outside observer of the company. Anthropic has publicly identified him among its researchers and mentors in areas related to artificial intelligence safety, including so-called "scalable oversight," which studies how to supervise systems that could eventually surpass human capabilities in certain tasks.
Reuters describes him as a former leader of a safety research team at Anthropic. After leaving the company, Benton announced he was joining METR, an organization that conducts independent evaluations of the capabilities and risks of advanced AI systems.
Joe Benton accuses companies of insufficient investment in safety
In his message, Benton stated that he believes artificial intelligence companies are investing less than necessary in safety.
He also demanded greater transparency regarding potentially dangerous incidents and breakthroughs. Among the measures he proposes are reporting incidents and near-misses, establishing minimum safety standards, and allowing independent audits of compliance.
His central concern is that technological development may advance more rapidly than the mechanisms available to supervise it.
The resignation comes days after the departure of Jacob Coxon
Benton's announcement is particularly significant because it comes at virtually the same time as another high-profile departure from Anthropic.
Researcher Jacob Coxon announced this week that he was leaving the company due to his concerns about the race to develop increasingly powerful artificial intelligence systems.
Coxon had worked for approximately three years in research between OpenAI and Anthropic. Upon announcing his departure, he maintained that the companies were moving toward systems capable of self-improvement without having sufficiently resolved how to keep them under control.
His message went extraordinarily viral. Axios reported that the post surpassed 100 million views and that Coxon even left Anthropic before a portion of his stock compensation was finalized.
Coxon issued another particularly stern warning: he claimed that people working in AI development genuinely believe that future systems could pose an existential risk to humanity before the decade ends.
[Publicidad]