key insights
- Safety teams will now participate earlier in model development and product launch decisions.
- OpenAI has consolidated research and safety leadership under a single executive.
- The changes arrive during an active period of product releases and executive departures.
OpenAI Safety Reorganization has prompted the departure of Johannes Heidecke, the company’s head of safety systems, as the artificial intelligence firm combines its safety and research organizations. The leadership overhaul places safety teams under a unified research structure as OpenAI accelerates frontier model development and responds to growing scrutiny over AI governance.
Heidecke told colleagues this week that he will be stepping down with the restructuring. In a memo, now seen by Wired, Chief Research Officer Mark Chen reported that Vice President of Research and Head of Alignment Mia Glaese will have a new role, encompassing both research and safety.
The leadership structure changes with the inclusion of safety in research
The restructuring is a major shift in OpenAI’s approach to safety oversight. Under the new structure, Glaese becomes vice president of research and safety, bringing together teams that previously operated under separate leadership.
Meanwhile, Saachi Jain will assume the role of interim head of safety systems and report directly to Glaese during the transition.
Chen told employees that integrating safety with frontier model development would allow safety experts to influence models and product launches much earlier. He also thanked Heidecke for his contributions to the company and emphasized that faster model releases require stronger coordination across engineering and safety teams.
Heidecke joined OpenAI as an AI safety analyst in 2021. In 2024, he became head of safety systems after Lilian Weng stepped down from the role. Weng later joined former OpenAI researchers in launching Thinking Machines Lab.
Leadership changes at a glance
Executive departures come during a period of rapid expansion
The OpenAI Safety Reorganization follows another senior leadership departure. Chief Futurist Joshua Achiam also informed colleagues this week that he would leave the company after nine years.
The changes arrive as OpenAI continues expanding its artificial intelligence products and commercial operations. The company recently introduced GPT-5.6, describing it as its strongest model for agentic programming while acknowledging that testing revealed more concerning misaligned behavior than earlier versions.
At the same time, leadership responsibilities have shifted elsewhere inside the company. OpenAI co-founder Greg Brockman will continue overseeing products and strategic initiatives after Fidji Simo stepped back because of a chronic illness. Simo remains involved as a part-time adviser following her medical leave.
The company has also increased its focus on consumer hardware. That effort gained momentum after OpenAI acquired IO Products, the startup founded by former Apple designer Jony Ive.
The need for AI governance keeps getting higher in the industry
The leadership shifts are part of a broader reshaping of the artificial intelligence sector as firms increasingly are under pressure to meet rising expectations for responsible AI development. With the development and increased use of frontier models, there has been a continuing push from governments, enterprise customers, and regulators for better governance.
By shifting the focus of safety closer to research, OpenAI appears to be investing more in embedding safety into the development lifecycle. As the competition and regulatory scrutiny on the AI industry grow, the restructuring could also have implications for other AI developers’ safety and governance strategies, as well as their research and development teams.
Industry pressure grows beyond internal restructuring
The OpenAI Safety Reorganization comes as governments and regulators continue examining how leading AI developers manage increasingly capable foundation models. Industry observers have pushed for stronger governance frameworks as companies release more advanced systems at a faster pace.
OpenAI’s latest restructuring reflects that changing environment. Chen told employees that frequent model training and shorter deployment schedules have increased coordination challenges across safety and engineering teams. Bringing both functions together aims to reduce delays while embedding safety reviews earlier in development.
The company also faces mounting legal and competitive pressure. Apple recently filed a federal lawsuit accusing OpenAI of using protected confidential information during its consumer hardware efforts. The conflict is a stark contrast to a previous attempt by the two firms in 2024 to bring ChatGPT into the iPhone ecosystem.
The legal proceedings have not been linked to the leadership changes but rather reflect the size of the pressures on OpenAI as it ventures into new areas.
Conclusion
As part of the OpenAI Safety Reorganization, Heidecke takes on a new role as the company experience a major leadership shift since the announcement. Research and safety will be brought under the wing of Mia Glaese, and OpenAI hopes to bring governance into the fold of the frontiers more tightly, while keeping product cycles quick.
The OpenAI Safety Reorganization also coincides with key product releases, a number of executive changes, legal action and regulatory scrutiny. These advancements underscore OpenAI’s evolving leadership dynamics in the rapidly competitive landscape of AI.
