Another leader of OpenAI’s safety strategy is leaving, joining a growing list of employees who have left the company’s safety and coordination teams in recent years.
Johannes Heidecke, head of OpenAI’s safety systems team, is leaving as the company reorganizes its safety and research activities under a single leader.
“We are grateful for Johannes’ contributions to OpenAI,” Mark Chen, OpenAI’s chief research officer, told Business Insider. “We look forward to the next chapter under Mia Grace’s leadership in research and safety.”
OpenAI said it will integrate safety more deeply across its research team under Glaese, who serves as vice president of research and safety.
The reorganization comes as OpenAI reconsiders the relationship between research and safety efforts. A spokesperson told Business Insider: “You can’t make good decisions about safety without understanding the functionality of the underlying model, and you can’t make good decisions about research without understanding the safety implications.”
OpenAI’s mission is to make artificial general intelligence a reality. — When we finally arrived — It benefits all humanity. “Safety, the practice of achieving positive impacts by mitigating the negative impacts of AI, is therefore core to our mission,” the company said.
But OpenAI has a poor track record of retaining top safety leaders, and some departing employees have publicly questioned the company’s commitment to the effort.
When Jan Leike, who previously co-led OpenAI’s superalignment team, retired in May 2024, he said the company was prioritizing product release over ensuring safety.
“In recent years, safety culture and processes have taken a backseat to shiny products,” Reich wrote in a post about X at the time. He said OpenAI needs to devote more resources to preparing future models and that building machines smarter than humans is an “inherently risky endeavor.”
Anthropic co-founder Dario Amodei served as OpenAI’s vice president of research until he resigned in 2020, citing disagreements over the company’s direction and approach to AI safety. A year later, he founded Anthropic as a safety-focused AI model maker.
Anthropic has been hit by departures of its own, most notably in February when Mlinanku Sharm, who led the safeguards research team, left the company and publicly warned that the AI company was facing pressure to “put aside what matters most.” Human safety researcher Dylan Scando also left this year to become OpenAI’s head of preparation.
Still, Anthropic’s core safety leadership remains largely intact compared to OpenAI. Renowned AI analyst Andrew Curran said that OpenAI’s ever-revolving safety door is a curse in relation to Harry Potter.
Meet eight leaders from OpenAI’s safety-focused teams who have retired in recent years.
Ilya Satskeva, co-founder
Sutskever, who co-founded OpenAI and served as its chief scientist for many years, left the company in May 2024. He co-led the company’s Superalignment initiative with Leike, a company founded to develop ways to control future superintelligent AI.
Upon leaving the company, he wrote in a post on X that OpenAI’s progress was “nothing short of miraculous.” He later established a specialized AI lab named Safe Superintelligence.
Jan Reik, Co-Founder
Laike left a few days after Sutskever. He offered one of the most direct public criticisms of OpenAI’s approach to safety, saying it had reached a “breaking point” after long disagreements with leadership.
After Leike and Sutskever left, OpenAI disbanded the Superalignment team and distributed its remaining employees to other research groups. Mr. Reich joined Anthropic to work on alignment research.
Miles Brundage, AGI Enabled Senior Advisor
Mr. Brundage retired from OpenAI in October 2024 after six years.
“OpenAI and the other frontier labs aren’t ready, and the world isn’t ready,” Brundage said at the time, referring to the impact AI could have on the world in his essay on Substack.
He wrote that staying with OpenAI was a tacit agreement with the company’s values. “Everyone working at OpenAI should take seriously the fact that what they do and say contributes to the organization’s culture, and that positive or negative path dependencies can arise as organizations begin to manage very sophisticated capabilities,” he wrote.
Lilian Wen, Vice President of Research and Safety
Wen, who previously led OpenAI’s safety systems team, left the company in November 2024 after about seven years, according to his LinkedIn profile.
She did not publicly raise concerns about OpenAI in her resignation statement. Instead, she praised the environment and said, “Training these models to be strong and responsible has set a new industry standard.”
Andrea Vallone, Model Policy Safety Research Director
Vallone retired at the end of 2025 after leading a team that shaped how OpenAI’s models respond in sensitive situations, such as conversations involving emotional dependency or mental health crises.
Her resignation was initially announced internally and she did not publicly criticize OpenAI. Vallone then joined Anthropic’s coordination team and continued to study the model’s behavior and safety.
Alexander Madrid, former Head of Preparation
Madry joined OpenAI to lead the Preparedness team, which assesses whether increasingly sophisticated models can create significant risks. He was removed from his leadership role in 2024 and shifted his focus to AI inference before retiring in May 2026.
Joshua Achiam, former Chief Futurist
Achiam retired in July. He previously led Mission Alignment, a group created to ensure the company’s AI development and deployment was aligned with the nonprofit’s mission.
“There’s no particular reason for me to leave, or a clear reason to leave now. But it’s something I’ve been thinking about for a while and it feels right. The world is in secret right now, and I feel it’s possible to work on the mission from outside the walls of Frontier Labs,” he wrote in a July post to X.
Johannes Heidecke, Head of Safety Systems
Heidecke joined OpenAI in 2021 and took over Safety Systems after Wen retired. His team helped evaluate and mitigate the risks of OpenAI models before they were released.
Heidecke has not made any public announcements since his retirement plans were made public on Friday. OpenAI said Heidecke decided to leave the company to more closely integrate safety operations with the broader research organization.