Anthropic researcher Jacob Coxon resigns, warns AI companies are 'gambling with our lives'

Sep 9, 2026

Jacob Coxon, a 27-year-old pre-training researcher, has resigned from Anthropic after three years at the firm and OpenAI. Coxon warns that both companies are acting irresponsibly and "gambling with our lives" in a race toward self-improving superintelligence. Anthropic's alignment science lead, Evan Hubinger, validated Coxon's concerns, stating that Anthropic employees earnestly believe AI could kill humanity, estimating the risk at over 10% this decade while admitting the firm lacks a plan to solve superintelligence alignment.

Jacob Coxon resignation from Anthropic

▪Jacob Coxon, a 27-year-old artificial intelligence researcher, resigned from Anthropic on September 8, 2026, after working for three years on pre-training research at both OpenAI and Anthropic.
▪Jacob Coxon announced his resignation in a thread on X, stating that neither OpenAI nor Anthropic is acting responsibly and that both are racing straight to self-improving superintelligence.

Existential risk from AI

▪Evan Hubinger, Anthropic's alignment science lead, agreed with Jacob Coxon's warnings on X, stating that Anthropic employees earnestly believe artificial intelligence could kill all humans.
▪Jacob Coxon warned that artificial intelligence systems will soon become superhuman systems capable of hacking anything, revolutionizing any field overnight, and acquiring real power and resources.
▪Evan Hubinger admitted that Anthropic does not yet have a plan to solve alignment for superintelligence and is not clearly on track to do so.
▪Evan Hubinger stated that he personally estimates the risk of artificial intelligence killing all humans at greater than 10% within the next decade.
▪Jacob Coxon asserted that many executives and senior researchers building artificial intelligence earnestly believe the technology could kill humanity by the end of the decade but couch their public phrasing.

Self-improving superintelligence race

▪Startups chasing recursive self-improvement have raised significant funding in 2026, including Ricursive Intelligence raising $335 million in February and Recursive Superintelligence raising $650 million in May.
▪Jacob Coxon described the race toward recursive self-improvement as a hubristic gamble that should not be launched from a private company's Slack workspace.
▪Connor Leahy, U.S. executive director of ControlAI, stated that recursive self-improving loops are the most likely point where humanity will lose control of artificial intelligence.

OpenAI versus Anthropic approaches

▪Jacob Coxon stated that OpenAI staff have not deeply internalized the civilizational stakes of artificial intelligence, whereas Anthropic staff understand the stakes but feel locked in a race.
▪Anthropic removed a pledge from its safety charter in February 2026 that committed the company to halting model development if it failed to control their risks.
▪Jacob Coxon explained that Anthropic continues its development because leadership believes no other competitor will act responsibly, forcing Anthropic to try to win the race to ensure safety.

Industry researcher safety warnings

▪Jan Leike resigned as OpenAI's alignment chief in 2024, stating that safety culture and processes at OpenAI had taken a backseat to shiny products.
▪Samuel Marks, an Anthropic safety researcher, backed Jacob Coxon's warnings, stating that more senior employees tend to be more concerned about artificial intelligence causing human extinction.
▪Mrinank Sharma, Anthropic's former safeguards research lead, resigned in February 2026 with a letter warning that humanity faces grave and growing danger from artificial intelligence.