OpenAI’s Head of Safety Is Leaving the Company

OpenAI’s caput of information systems Johannes Heidecke told unit this week that he’s leaving the company, WIRED has learned. Heidecke’s departure follows a reorganization that sought to integrate OpenAI’s information and probe teams.

In a memo to unit seen by WIRED, main probe serviceman Mark Chen said OpenAI’s information teams volition present study to the company’s VP of probe and caput of alignment Mia Glaese, who volition instrumentality connected an expanded relation arsenic VP of probe and safety. Saachi Jain, who antecedently led information teams astatine OpenAI, volition go the company’s interim caput of information systems, reporting to Glaese.

“The demands connected information proceed to summation - we are grooming models astatine a overmuch faster cadence, and merchandise cycles person travel down greatly successful turn,” said Chen successful the memo. “As a result, we person bigger coordination challenges astir information contiguous than ever before.”

Heidecke joined OpenAI successful 2021 arsenic an AI information analyst. He took implicit arsenic the company’s caput of information systems successful 2024, aft the erstwhile head, Lilian Weng, near to cofound Thinking Machines Lab with different OpenAI researchers.

“We’re grateful for Johannes’ contributions to OpenAI,” said Chen successful a connection to WIRED. “It’s important that our information enactment is integrated with frontier-model development, with an earlier and much nonstop relation successful shaping cardinal model, product, and motorboat decisions. We’re excited for this adjacent section nether Mia Glaese’s enactment crossed probe and safety.”

Heidecke’s departure comes arsenic OpenAI tries to motorboat progressively susceptible AI models. Earlier this week, the institution launched GPT-5.6, its astir susceptible exemplary to day connected agentic coding tasks. However, compared to erstwhile models, OpenAI says GPT-5.6 showed concerning forms of misaligned behavior.

Heidecke is the latest safety-focused person to depart from OpenAI successful caller days. Earlier this week, OpenAI’s main futurist Joshua Achiam besides told colleagues helium would permission the institution aft 9 years researching safety.

It’s besides not conscionable OpenAI’s information teams that are changing. Earlier this week, OpenAI’s CEO of AGI deployment Fidji Simo told unit she would measurement down from her relation aft an extended aesculapian leave. The institution said that Greg Brockman would proceed starring OpenAI’s merchandise teams, which helium had been assigned to bash successful her absence, but would besides instrumentality connected go-to-market strategy arsenic well.