OpenAI's Top Safety Executive to Depart July 24 in IPO Run-Up

Another senior safety leader is on the way out at OpenAI, and this one lands at an awkward moment. Wired's reporting has one of the company's top safety and policy executives leaving on July 24, adding to a run of departures as OpenAI moves toward an initial public offering. Johannes Heidecke has been running the company's safety systems team, and briefly inherited former model-policy chief Andrea Vallone's team after her late-2025 exit to Anthropic.

On its own the news would be a single leadership change, but the pattern is what makes it worth paying attention to. Reporting on this month's earlier exit of chief futurist Joshua Achiam described his departure as 'at least the fifth senior safety-focused leader to leave OpenAI in a roughly two-year span,' following Jan Leike, Miles Brundage, Steven Adler and Andrea Vallone. Leike and Vallone went to Anthropic. Brundage and Adler left in 2024 to start nonprofits pushing labs on safety and security standards. Add the February 2026 dissolution of OpenAI's 16-month-old Mission Alignment unit and the picture is one where safety leadership has been steadily thinning and getting reorganized into other functions.

Why this matters if you are not tracking OpenAI's org chart: this is happening while the company is publicly preparing for an IPO. Governance and safety oversight are exactly the details institutional investors and regulators read closely, and rapid safety-leadership churn is the kind of soft signal that ends up in prospectus risk factors and in regulator questions, whether or not the underlying safety work is intact.

The honest caveat is that departures are not the same as an abandoned safety program, and the reporting available so far does not spell out a stated reason for this specific exit. What it also does not yet give you is a named permanent successor, a clear new reporting line for the Preparedness Framework, or a disclosure of how the head-of-safety remit will be governed through the IPO transition. Those are the concrete details worth watching over the next few weeks.

The upside, if you follow the people rather than the org chart, is real. Ex-OpenAI safety leaders have been strengthening Anthropic's roster and standing up nonprofits that press frontier labs on safety and security standards. If OpenAI's internal accountability structure keeps thinning, more of the actual check on how these systems get deployed shifts to those outside actors, and to whoever OpenAI decides to name next.