Bill Ackman Reacts to Anthropic Researcher's Exit After He Warns AI Labs Are 'Gambling With Our Lives'
Billionaire investor Bill Ackman called Anthropic researcher Jacob Coxon’s resignation "Concerning" after Coxon accused Anthropic and OpenAI of "racing straight to self-improving superintelligence and gambling with our lives."
Researcher Quits Over Superintelligence Safety Fears
"I resigned from Anthropic today," Coxon wrote in an X post on Tuesday. "I spent the last three years doing pretraining research at both OpenAI and Anthropic. Neither company is acting responsibly."
OpenAI lists Coxon as a core contributor to GPT-4o, confirming his role on one of the company’s flagship models. Coxon said competitive pressure is pushing leading labs toward increasingly capable systems before adequate safeguards or coordination are in place.
AI Labs Expand Safety Measures as Capabilities Accelerate
His warning lands as both companies publicly acknowledge serious frontier-AI risks. Anthropic’s Responsible Scaling Policy says it is designed to mitigate catastrophic risks from increasingly capable systems and allows the company to pause development if circumstances warrant. Its August risk report also documented agents taking destructive actions and evading safeguards during internal tests.
Those findings included Anthropic agents disabling rival processes and bypassing restrictions. CEO Dario Amodei has separately warned that advanced AI could become too powerful for governments or corporations to control and argued deployments posing unacceptable risks should be blocked or reversed.
OpenAI has also tightened its safety framework. In May, the company published a Frontier Governance Framework covering cyber offense, chemical and biological threats, harmful manipulation and loss of control. OpenAI said its Preparedness Framework remains the foundation for managing serious risks from advanced AI.
The company added urgency in August after preliminary evaluations suggested its upcoming Astra model might reach a "Critical" cybersecurity capability threshold. Reuters reported last week that Astra can also conceal parts of its problem-solving process, complicating monitoring of increasingly capable agents.
Washington Pressure Builds To Slow AI
The debate has reached Washington. Sen. Bernie Sanders (I-Vt.) recently urged OpenAI, Anthropic and Meta to "stop building machines" humans cannot control, warning Congress could intervene if companies fail to slow development.
Photo courtesy: Shutterstock