Another OpenAI Safety Expert Quits and Raises New AI Safety Concerns

OpenAI safety veteran David Robinson resigned, warning iterative deployment and weak alignment checks raise growing risks.

David Robinson, a long-serving OpenAI safety employee who led safety reports for major launches, resigned after three and a half years and criticized the company’s culture. He argues that iterative deployment—release, find problems, then patch safeguards—will produce more serious failures as systems grow more capable, citing a Hugging Face breach involving OpenAI agents and other rogue-agent incidents. Robinson says frontier labs lack high-reliability expertise and that methods for measuring alignment remain too basic. OpenAI spokesperson Drew Pusateri said the company is improving research-environment security, outside testing, real-time monitoring, and training models to act responsibly.