OpenAI safety and transparency leader David Robinson resigns, warns of rapid launches
Robinson said AI firms should operate like nuclear power plants, with multiple layers of safety checks
Published
David Robinson, who led safety and transparency work at OpenAI, has resigned, saying the company is launching increasingly capable AI systems faster than it can safeguard them.
Robinson said AI firms should operate more like nuclear power plants, built with multiple layers of safety rather than a single check. OpenAI said in response that it is expanding third-party evaluations, security testing and real-time monitoring of models it considers potentially dangerous.
The resignation follows a string of disclosures OpenAI has made this year, including voluntary reports of breaches and training pauses through September. California's attorney general escalated that scrutiny on October 1 with a subpoena over cybersecurity incidents, a compulsory step beyond what OpenAI had disclosed on its own. Separate reporting has said Sam Altman is not closely involved in security matters at the company.
Robinson's departure adds a named, high-profile voice to concerns that have so far coexisted with OpenAI's financial momentum. The company closed a funding round on October 1 in which SoftBank and Nvidia each completed $10 billion investments, finishing $30 billion pledges apiece for $20 billion combined, as OpenAI's annualized revenue approaches $70 billion and its pre-IPO valuation stands at $1.2 trillion. The warning sharpens the question of whether regulators or OpenAI itself forces a slowdown before those growth figures face further scrutiny.