OpenAI Safety Chief Exits Amid Restructure

Johannes Heidecke, OpenAI's head of safety systems, is leaving the company following a reorganization that merges safety and research teams , according to Wired . Saachi Jain, who has led OpenAI's safety teams before, will slot in as interim head of safety systems .

Heidecke informed colleagues this week that he will leave the company by July 24 , having first started at OpenAI in 2021 . Chief research officer Mark Chen said OpenAI's safety teams would report to vice president of research and head of alignment Mia Glaese, whose role has been expanded to oversee both research and safety .

Why Fold Safety Into Research?

OpenAI's chief research officer, Mark Chen, told Wired in a statement that it was "important that our safety work is integrated with frontier-model development, with an earlier and more direct role in shaping key model, product and launch decisions" .

Chen wrote in a memo that "the demands on safety continue to increase—we are training models at a much faster cadence, and release cycles have come down greatly in turn" . He added, "As a result, we have bigger coordination challenges around safety today than ever before" .

Instead of safety operating as its own distinct pillar, it's being folded into the research organization . Chief Research Officer Mark Chen told staff in a memo that safety teams would now report to Mia Glaese, whose title has been expanded to VP of Research and Safety .

The Sixth Safety Exit in Two Years

Heidecke is the sixth senior safety-focused leader to leave OpenAI in approximately two years . Before him: Jan Leike, who co-led the Superalignment team and left for Anthropic; Miles Brundage, who left to build an external AI safety nonprofit; Steven Adler, who did the same; Andrea Vallone, who joined Leike at Anthropic at the end of 2025; and Chief Futurist Joshua Achiam, who notified colleagues earlier this month that he would depart after nine years .

The attrition traces its roots to May 2024, when OpenAI's Superalignment team—co-led by co-founder Ilya Sutskever and Leike, and backed by a pledge of 20% of the company's computing resources—effectively dissolved following both leaders' resignations . Leike wrote on X that OpenAI's "safety culture and processes have taken a backseat to shiny products" .

Joshua Achiam, OpenAI's chief futurist, told colleagues on Tuesday that he would depart after nearly nine years at the company, saying the move was not prompted by a specific reason, but had been on his mind for some time . OpenAI disbanded the mission alignment team in February, when Achiam moved into the chief futurist role .

Timing Alongside GPT-5.6 Launch

The staff shifts come on the heels of OpenAI's latest model release, GPT-5.6, after it was recently approved by the US government . GPT-5.6 is a large language model developed and released by OpenAI on July 9, 2026 for the public .

Initially it was supposed to be released in the month of June, but due to government restrictions levied by the Trump Administration it was only available as a limited preview on June 26, 2026 . At the request of the U.S. government, OpenAI shipped GPT-5.6 first to a small group of trusted partners, gated behind a government safety review .

Under OpenAI's Preparedness Framework, all three members of the GPT-5.6 family are treated as having "High capability" in both the Cybersecurity and Biological/Chemical risk domains . When GPT-5.6 Sol crossed the framework's "High" risk classification for cybersecurity capabilities, OpenAI coordinated with the White House on a phased release—initially limiting access to approximately 20 government-approved partners before the model opened to the public this week .

Does Embedded Safety Work?

Critics argue the reorganization eliminates the independent authority that makes safety oversight effective—the same argument applies in aviation, nuclear energy, and pharmaceutical industries, where independent oversight bodies exist precisely because embedded safety functions tend to lose authority over the commercial functions they are supposed to check .

The decision to merge safety under the research umbrella can be read two ways: the optimistic interpretation is that safety becomes deeply embedded in everything the research team builds, rather than operating as an afterthought or a separate checkpoint; the skeptical interpretation is that safety loses its independent voice and becomes subordinate to the team whose primary incentive is to ship products faster .

Heidecke is the sixth senior safety-focused leader to leave OpenAI in approximately two years . His departure arrives four days after the Future of Life Institute gave OpenAI a C on its Summer 2026 AI Safety Index—a score that FLI's panel described as evidence of an industry that has "moved the goalposts" on its own prior safety commitments .

What Changed This Week

OpenAI eliminated the independent organizational standing of its safety function, folding safety teams under research leadership in a restructuring that costs the company its head of safety systems. Heidecke's departure extends a two-year pattern of safety-leadership exits that critics say has left the company without a genuinely independent check on its own development decisions. The reorganization means safety teams now report to Mia Glaese, whose existing role as vice president of research and head of alignment has been expanded into a new title: vice president of research and safety.

What to Watch

Saachi Jain has been named interim head of safety systems while the company searches for Heidecke's permanent replacement . Whether OpenAI fills the role or quietly retires it will signal how the company balances safety oversight with commercial pressure. The restructuring comes as OpenAI continues to expand its AI capabilities while facing increased scrutiny from regulators, governments and industry observers over AI safety, governance and the development of increasingly powerful frontier models . Forty-two state attorneys general have opened an investigation into the company, serving a subpoena on advertising, user data, and internal policies shortly after it confidentially filed for a stock market listing .

Reporting based on coverage from Bloomberg, Wired, Engadget, TechTimes, The Next Web, Gizmodo, July 11-12, 2026.