AI Whistleblowers Warn About Unchecked, Unsafe Development at OpenAI and Anthropic

Handshake of man and robot. Modern technologies. Art collage.

Last week, an AI whistleblower warned via X that major AI development companies OpenAI and Anthropic are racing to build superintelligence systems as quickly as possible, without building sufficient safeguards to prevent those systems from making independent and actively harmful decisions.

Jacob Coxon, a former researcher at both OpenAI and Anthropic, announced his resignation from the latter company on September 8.

“Neither company is acting responsibly,” his post said. “They are racing straight to self-improving superintelligence and gambling with our lives.”

Coxon also stated that while AI companies publicly report that the dangers posed by new AI systems are relatively minimal, “the same people express fear privately.” He went so far as to state: “The people building AI earnestly believe that it could kill us all by the end of the decade.”

In a recent statement on AI safety concerns, also on X, OpenAI CEO Sam Altman stated that safety procedures should be increased to ensure new models do not exhibit undesirable behavior, but maintained that companies like his are capable of implementing such safeguards on their own. “When we talk about “pacing”, he further clarified, “we do not mean “stopping”. Progress has been rapid and will continue to be.”

Whistleblower advocates like attorney and National Whistleblower Center chairman Stephen M. Kohn argue that disclosures like Coxon’s demonstrate a need to protect AI whistleblowers and ensure this type of information reaches the public. They assert that disparities between these companies’ external confidence and internal concerns, combined with a culture of secrecy that leaves many insiders too scared to blow the whistle, puts the public in danger.

The AI Whistleblower Protection Act, a bipartisan bill introduced by Senator Chuck Grassley (R-IA) in 2025, is one potential mechanism to increase whistleblower disclosures from AI insiders. It would provide anti-retaliation protections and due process to AI whistleblowers who report unethical behavior or undisclosed risks internally or to the government.

“Coxon’s courage is admirable,” Kohn said. “But whistleblowers should not have to risk their careers to come forward.”

Exit mobile version