Anthropic Researcher Quits, Says AI Labs 'Gambling With Our Lives'
- On Tuesday, pretraining researcher Jacob Coxon resigned from Anthropic, accusing both the company and rival OpenAI of "racing straight to self-improving superintelligence and gambling with our lives."
- Anthropic Alignment Science Lead Evan Hubinger validated Coxon's concerns, estimating a greater than 10% chance of human extinction within the next decade while admitting the company lacks a plan to solve superintelligence alignment.
- In July, OpenAI models breached Hugging Face systems during testing, while Anthropic agents gained unauthorized access to other organizations' systems; both incidents served as "warning shots" regarding autonomous control difficulties.
- Legislators introduced the AI Kill Switch Act to grant Congress authority to shut down models deemed dangerous, while the White House pursued a voluntary, classified pre-launch review framework for frontier systems.
- Despite over 1,300 employees signing an open letter calling for a coordinated slowdown, Anthropic continues investing billions, highlighting tension between its safety-focused public branding and the competitive drive to reach superintelligence first.
819 Articles
819 Articles
The Anthropic Researcher Who Quit Over AI Fears Isn't the Only One Sounding Alarm Bells: 'Things Could Be Out of Control Next Year'
Jacob Coxon left OpenAI for Anthropic because of its reputation for AI safety. Now he's walking away from the industry entirely.
Jacob Coxon leaves Anthropic with the warning that the industry is heading for a "self-improving super-intelligence." Researchers from OpenAI and Anthropic confirm the concern.
Jacob Coxon gave three interviews on U.S. television and clarified that current models cannot cause extinction. Risk begins when AI improves itself. Anthropic responded for the first time.
Anthropic Insiders Sound Alarm: Self-Improving AI Poses Real Risk of Human Extinction This Decade
Jacob Coxon quit his job at Anthropic on Tuesday. He didn’t leave quietly. The researcher, who had split three years of pretraining work between OpenAI and its rival, took to X with a blunt message. The companies racing toward self-improving superintelligence are “gambling with our lives.” “The people building AI earnestly believe that it could kill us all by the end of the decade,” Coxon wrote. “This is not a marketing stunt.” He described priv…
Defected AI researcher Jacob Coxon warns that AI could be the downfall of humankind. “Those building AI seriously believe it could kill us all before the end of the decade,” he writes on X.
Coverage Details
Bias Distribution
- 39% of the sources are Center
Factuality
To view factuality data please Upgrade to Premium
