Anthropic Researcher Resigns, Warning That AI 'Could Kill Us All By The End Of The Decade'

Summary

Former Anthropic and OpenAI researcher Jacob Coxon resigned, warning that frontier artificial intelligence labs are racing recklessly toward dangerous superintelligence. His concerns were echoed by colleagues who cited non-trivial chances of human extinction, even as United States officials resist development pauses over competitive fears.

Key Takeaways

1. A researcher for Anthropic resigned from the company on Tuesday, warning that “the people building AI earnestly believe that it could kill us all by the end of the decade.”
2. Evan Hubinger, alignment science lead at Anthropic, said Coxon’s assessment was valid, adding that he believes there is an over 10% chance the technology could wipe out all humans.
3. “We can’t pause. You can’t, because the Chinese won’t pause. Even the North Koreans, they won’t pause,” Treasury Secretary Scott Bessent said Tuesday.

Statement Breakdown

• Claimed Facts: 50% of statements the article presents as facts
• Opinions: 35% of statements classified as editorial or subjective
• Claims: 15% of statements surfaced for additional reader evaluation

Credibility & Bias Reasoning

Credibility assessment: The reporting accurately documents public statements, social media posts, and government remarks from named figures in artificial intelligence. It cross-references internal company disclosures and previous safety declarations. Some assertions rely heavily on subjective probability estimates regarding human extinction, which inherently lack empirical verification.

Bias assessment: Alarmist AI Safety Lens. The article focuses primarily on catastrophic risk perspectives from safety researchers and critics. It highlights worst-case existential scenarios while framing government reluctance to regulate purely as geopolitical rivalry. Opposing arguments emphasizing current technical limitations receive minimal attention.

Note: Contains subjective assessments of future technological catastrophe alongside confirmed resignations and corporate disclosures.

Credibility flag: Existential Risk Warnings

Claimed Facts (5)

• Reports the specific employment history and communication channel of the researcher.
• States a clear, checkable position expressed publicly on social media regarding current model safety.
• Documents a verified historical public statement signed by industry executives in 2023.
• Cites a specific formal disclosure by Anthropic regarding model penetration tests.
• References verifiable reporting by the Financial Times concerning regulatory testing access.

Opinions (4)

• Expresses a subjective value judgment regarding corporate motivations and risk tolerance.
• Articulates a personal concern and future threat model.
• Presents an interpretive projection about societal impact during the technology transition.
• Reflects a subjective assessment of societal and governmental preparedness.

Claims (5)

• Makes extreme, sweeping assertions about near-term omnipotent hacking capabilities without empirical proof.
• Generalizes the beliefs of an entire industry while asserting an absolute existential ranking.
• Presents a speculative probability figure for human extinction lacking testable scientific backing.
• Attributes a single catastrophic belief to all developers across the sector.
• Speculates on a multi-year extinction horizon based on informal peer impressions.

Key Sources

• Jacob Coxon — Former Pretraining Researcher at Anthropic and OpenAI
• Evan Hubinger — Alignment Science Lead at Anthropic
• Samuel Marks — Scalable Oversight Lead at Anthropic
• Bill Gates — Co-founder of Microsoft
• Scott Bessent — United States Treasury Secretary
• Financial Times — Financial and Business Daily Newspaper
• Marita Vlachou — Reporter at HuffPost