Anthropic Safety Researcher Warns AI Carries 10 Percent Doom Risk

The commercial pressure to build advanced artificial intelligence has triggered a fresh wave of internal alarm at top laboratories, exposing deep cracks between commercial growth and existential safety research. A senior safety researcher at San Francisco-based artificial intelligence lab Anthropic has publicly stated that there is a greater than 10 percent probability that advanced artificial intelligence could cause human extinction before 2030. This stark warning arrived mere hours after a fellow researcher walked away from the company, citing a reckless industry-wide sprint toward uncontrollable superhuman intelligence systems. The public disclosures lay bare the growing tension inside premier frontier labs, where elite engineers are beginning to question the very commercial incentives they help sustain.

Why Jacob Coxon Resigned From Anthropic

Jacob Coxon, an artificial intelligence researcher who previously trained foundational systems at OpenAI before moving to Anthropic, announced his resignation in September 2026 via a post on the social media platform X. Coxon explicitly accused both Anthropic and OpenAI of gambling with human lives by accelerating directly toward self-improving superintelligence without adequate safety guardrails. According to The Verge, Coxon resigned specifically over what he characterized as a lax internal approach to managing existential risk. His departure marks another high-profile exit from the artificial intelligence sector, following a pattern where technical safety talent walks away when commercial deployment timelines outpace theoretical alignment research.

The convergence of Coxon's departure and the 10 percent existential risk estimate underscores a recurring structural vulnerability in modern artificial intelligence development. Labs are racing to commercialize models while internal safety teams grapple with failure modes they cannot yet reliably predict or control.

• Jacob Coxon resigned from Anthropic over safety concerns related to the rapid race for superintelligence.
• The departing researcher previously trained AI systems at OpenAI before joining Anthropic.
• A senior Anthropic safety researcher separately calculated a greater than 10 percent chance of human extinction from AI by 2030.

The Growing Debate Over Superintelligence Risks

The artificial intelligence industry faces unprecedented scrutiny regarding existential risk as labs cross critical development milestones toward autonomous self-improvement. Critics argue that market competition between major players leaves little room for rigorous, unhurried safety evaluations, forcing researchers to choose between corporate velocity and ethical caution. While executives at labs like Anthropic and OpenAI frequently emphasize their commitment to responsible scaling, internal dissent suggests that voluntary frameworks are failing to satisfy safety-minded engineers. This friction threatens to trigger a brain drain of top-tier alignment talent away from commercial labs and toward independent research institutions or regulatory bodies.

"Racing straight to self-improving superintelligence and gambling with our lives." — Jacob Coxon, Former Anthropic and OpenAI Researcher

What to Watch Next

Industry observers and enterprise buyers should monitor three critical indicators to gauge how artificial intelligence labs are handling internal safety dissent and regulatory pressure:

• Whether additional safety researchers follow Jacob Coxon in resigning from Anthropic or competitor labs over scaling velocity.
• How executive leadership at Anthropic and OpenAI publicly respond to mounting internal and external criticism regarding self-improving superintelligence.
• Any shifts in regulatory oversight or voluntary safety commitments following high-profile whistleblower departures in the artificial intelligence sector.