Johannes Heidecke, OpenAI’s head of safety systems since 2021, is leaving the company as part of a restructuring of its safety and research teams. Engadget reported the departure on Saturday, citing a memo to staff obtained by Wired in which Heidecke announced his exit. The move puts a familiar question back on the table: who at OpenAI owns the authority to slow a launch, and does that authority survive a reorg intact.

Saachi Jain, who has previously led OpenAI’s safety teams, will step in as interim head of safety systems, according to the Wired report cited by Engadget. The word interim matters here. Naming an interim leader defers the decision about permanent ownership of the safety function rather than resolving it, and it leaves the org chart unsettled at the exact moment a new reporting structure is taking shape.

That new structure routes safety through Mia Glaese, who becomes OpenAI’s vice president of research and safety. Folding safety into a joint research and safety title is a structural choice with real tradeoffs. It can mean safety review happens earlier, before a model ships rather than after. It can also mean safety loses a dedicated executive whose only mandate is to say no to a launch.

Mark Chen, OpenAI’s chief research officer, told Wired that the goal is for safety work to be “integrated with frontier-model development, with an earlier and more direct role in shaping key model, product and launch decisions.” That is a claim about process, not an outcome. Whether an earlier seat at the table translates into more leverage over launch timing, or simply more visibility without more veto power, will not be testable until the next contested release.

Safety leadership churn is not new at OpenAI. The pattern of safety executives departing and reporting lines being redrawn has recurred enough times that a single exit reads less as an isolated resignation and more as evidence of an unresolved structural question inside the company. Each reorg resets who has to re-earn credibility with the safety-focused engineers who remain.

The timing compounds the stakes. This shift lands right after OpenAI shipped GPT-5.6, which the company says has cleared US government approval, and while OpenAI is reportedly consolidating decision-making power ahead of a planned IPO. A public listing typically pulls authority toward executives who answer to shareholders and boards, not toward standalone risk functions. Where safety sits in that chart, as a peer function with launch veto or as a subordinate input to research, is precisely what an interim appointment leaves open.

OpenAI still lists a Head of Preparedness, a role Sam Altman said on X was created earlier this year to “prepare for and mitigate … severe risks.” That role remains intact, which suggests the frontier-risk function is not being touched even as the safety systems reporting line is redrawn. The distinction is worth tracking: preparedness and safety systems cover different failure modes, and a reorg that reshuffles one without disturbing the other tells you which risks the company currently treats as most acute.

For anyone tracking OpenAI’s governance ahead of an IPO, the next signal to watch is who gets named permanent head of safety systems, and whether that person reports at the same level Heidecke did or one rung lower under Glaese. That single reporting-line detail will say more about OpenAI’s actual safety priorities than any statement from Chen or Altman.

Engadget (Jackson Chen) reported the departure on July 11, 2026, citing Wired’s review of an internal staff memo.