Another leader of OpenAI’s safety strategy is leaving, joining an ever-growing list of employees who have left the company’s safety and alignment teams in recent years.
Johannes Heidecke, the head of OpenAI’s Safety Systems team, is departing as the company reorganizes its safety and research work under a single leader.
“We’re grateful for Johannes’ contributions to OpenAI,” Mark Chen, OpenAI’s chief research officer, told Business Insider. “We’re excited for this next chapter under Mia Glaese’s leadership across research and safety.”
OpenAI said it would integrate safety more deeply across its research teams under Glaese, who will now serve as its vice president of research and safety.
The reorganization comes as OpenAI reconsiders the relationship between its research and safety efforts. A spokesperson told Business Insider that “you can’t make good safety decisions without understanding the underlying model capabilities, and you can’t make good research decisions without understanding the safety implications.”
OpenAI’s mission is to ensure that artificial general intelligence — when it finally arrives — benefits all of humanity. “Safety — the practice of enabling AI’s positive impacts by mitigating the negative ones — is thus core to our mission,” the company says.
OpenAI, however, has a poor track record of retaining its top safety leaders, and some departing employees have publicly questioned its commitment to that work.
When Jan Leike, who once co-led OpenAI’s superalignment team, left in May 2024, he said the company was prioritizing releasing products over ensuring their safety.
“Over the past years, safety culture and processes have taken a backseat to shiny products,” Leike wrote in a post on X at the time. He said OpenAI needed to devote more resources to preparing for future models and that building machines smarter than humans was an “inherently dangerous endeavor.”
Anthropic cofounder Dario Amodei was OpenAI’s vice president of research before he left in 2020 over disagreements about the company’s direction and approach to AI safety. He started Anthropic a year later as an AI model maker committed to safety.
Anthropic has suffered its own departures, notably in February when Mrinank Sharm, who led its Safeguards Research Team, left and publicly warned that AI companies face pressure to “set aside what matters most.” Dylan Scand, an Anthropic safety researcher, also left this year to become OpenAI’s head of preparedness.
Still, Anthropic’s core safety leadership has largely remained intact compared to OpenAI. Andrew Curran, a prominent AI analyst, called OpenAI’s ever-revolving door of safety leads a curse in reference to Harry Potter.
Here are eight leaders of OpenAI’s safety-focused teams who have left in recent years.
Ilya Sutskever, cofounder
Sutskever, an OpenAI cofounder and its longtime chief scientist, left in May 2024. He had co-led the company’s Superalignment initiative with Leike, which was created to develop methods for controlling future superintelligent AI.
When he left, he said OpenAI’s progress had been “nothing short of miraculous,” in a post on X. He later founded Safe Superintelligence, an AI lab dedicated solely to its namesake.
Jan Leike, cofounder
Leike left days after Sutskever. He offered one of the most direct public criticisms of OpenAI’s approach to safety, saying he had reached a “breaking point” after prolonged disagreements with leadership.
OpenAI dissolved the Superalignment team after Leike and Sutskever left and distributed its remaining employees across other research groups. Leike joined Anthropic to work on alignment research.
Miles Brundage, senior advisor for AGI readiness
Brundage left OpenAI in October 2024 after six years.
“Neither OpenAI nor any other frontier lab is ready, and the world is also not ready,” Brundage wrote in an essay on his Substack at the time, referring to the impact AI could have on the world.
He wrote that staying at OpenAI is an implicit agreement with the company’s values. “Anyone working at OpenAI should take seriously the fact that their actions and statements contribute to the culture of the organization and may create positive or negative path dependencies as the organization begins to steward extremely advanced capabilities,” he wrote.
Lilian Weng, vice president of research and safety
Weng, who previously led OpenAI’s Safety Systems team, left the company in November 2024 after almost seven years, according to her LinkedIn profile.
She did not publicly raise concerns about OpenAI in her departure statement. Instead, she praised the environment, saying that it’s work in “training these models to be both powerful and responsible has set new industry standards.”
Andrea Vallone, head of model-policy safety research
Vallone left at the end of 2025 after leading the team that shaped how OpenAI’s models respond in sensitive situations, including conversations involving emotional dependence and mental-health crises.
Her departure was initially announced internally, and she did not publicly criticize OpenAI. Vallone later joined Anthropic’s alignment team to continue studying model behavior and safety.
Aleksander Madry, former head of preparedness
Madry joined OpenAI to lead its Preparedness team, which evaluates whether increasingly capable models could create severe risks. He was moved out of that leadership role in 2024 and shifted his attention to AI reasoning before leaving the company in May 2026.
Joshua Achiam, former chief futurist
Achiam left in July. He previously led Mission Alignment, a group created to ensure the company’s AI development and deployment align with its nonprofit mission.
“There’s not a specific reason for me leaving, or a specific reason for why now. But it’s something I have been thinking of for a while and it feels right. The world is in on the secret now and it feels possible to work on the mission from outside the walls of a frontier lab,” he wrote in a post on X in July.
Johannes Heidecke, head of Safety Systems
Heidecke joined OpenAI in 2021 and took over Safety Systems after Weng’s departure. His team helped evaluate and mitigate risks in OpenAI’s models before their release.
Heidecke has made no public statements since his planned departure became public on Friday. OpenAI said Heidecke decided to leave as the company integrated safety work more closely with its broader research organization.
Read the full article here
