OpenAI integrates safety into its research as the head of safety departs.

      Johannes Heidecke, OpenAI’s head of safety systems, is departing the company due to an internal reorganization that combines the safety and research teams under a single leader, as reported by Wired on Friday. Chief Research Officer Mark Chen informed staff in a memo that the safety teams will now report to Mia Glaese, who has been promoted to the newly created position of VP of Research and Safety. Saachi Jain has been appointed as the interim head of safety systems while the company searches for a permanent replacement for Heidecke. This marks the second time in under two years that OpenAI has integrated its safety organization into a structure led by a research head.

      Heidecke's Background

      Heidecke began his tenure at OpenAI in 2021 as an AI safety analyst and became the head of safety systems in 2024, taking over from Lilian Weng. His responsibilities included work on model alignment, rule-based reward systems, and evaluating the company's preparedness for potentially risky model capabilities. Chen expressed gratitude to Heidecke in his communication to employees, emphasizing the importance of integrating safety efforts with frontier-model development and ensuring that it plays a proactive role in crucial model, product, and launch decisions. Heidecke is the latest in a series of senior safety personnel to leave or be reassigned within the company over the past two years.

      A Trend of Disbandments

      OpenAI's Superalignment team, introduced in 2023 with a commitment of 20% of the company’s computing resources, was disbanded in May 2024 after the departure of its co-leads, Ilya Sutskever and Jan Leike. Leike publicly stated upon his exit that "safety culture and processes have taken a backseat to shiny products." The AGI Readiness team was dissolved in October 2024 when its leader Miles Brundage resigned, followed by the disbandment of the Mission Alignment team, which was Superalignment’s successor, in February 2026 after 16 months. Its leader, Joshua Achiam, was given the new title of “chief futurist.” In April, OpenAI also experienced a significant loss of leadership, with the resignation of its product chief, the head of Sora, and its enterprise CTO on the same day.

      Heidecke's exit coincides with a trend of senior leaders leaving the company, including Fidji Simo, OpenAI’s chief of applications, who stepped down this month for an extended medical recovery.

      The Case for Integration

      Glaese’s expanded title indicates OpenAI’s intention to maintain safety as a clear priority within the new organizational structure. On April 6, the company launched a Safety Fellowship, inviting external researchers to engage in independent safety and alignment work at the lab. Chen's rationale is that integrating safety within research ensures its involvement in model decisions from the outset rather than serving merely as a final checkpoint before launch. Some critics contend that a safety team embedded within research lacks the structural independence and authority necessary to delay or block products compared to a team that operates separately.

      External Pressures

      This departure occurs amidst increasing external scrutiny faced by OpenAI. Forty-two state attorneys general have initiated an investigation into the company, serving a subpoena regarding advertising, user data, and internal policies shortly after it confidentially applied for a stock market listing. Lilian Weng, Heidecke’s predecessor in the safety systems role, has transitioned to Thinking Machines Labs, the AI startup established by former OpenAI CTO Mira Murati, who has publicly warned that AI governance is not keeping pace with the capabilities of models.