Jan Leike, a leading AI researcher who recently left OpenAI, has joined Anthropic to lead a new superalignment team. The move puts one of the most visible names in technical AI safety inside a company that has often presented itself as more safety-focused than OpenAI.
A prominent AI safety researcher changes teams
Leike resigned from OpenAI earlier this month before publicly criticizing the company’s approach to AI safety. He has now moved to Anthropic, an OpenAI rival, where his work will continue under the banner of superalignment.
In a post on X, Leike said his new team at Anthropic will focus on "scalable oversight," "weak-to-strong generalization" and automated alignment research. He also wrote that he was excited to continue the superalignment mission and invited interested candidates to reach out.
The hiring is notable because Leike previously co-led OpenAI’s Superalignment team. That team had aimed to solve the core technical challenges of controlling superintelligent AI in the next four years, but it was recently dissolved.
What the new Anthropic team will study
The work described by Leike sits in a technical area of AI safety concerned with whether powerful AI systems can be guided, checked and kept aligned with intended behavior. The source article names three focus areas for the new team:
- Scalable oversight: techniques to control large-scale AI’s behavior in predictable and desirable ways.
- Weak-to-strong generalization: one of the areas Leike said the team will work on.
- Automated alignment research: another focus Leike identified for the team.
A source familiar with the matter told TechCrunch that Leike will report directly to Jared Kaplan, Anthropic’s chief science officer. The same source said Anthropic researchers already working on scalable oversight will move to report to Leike as the new team spins up.
That reporting line matters because it places the new group close to Anthropic’s scientific leadership. It also suggests the company is not treating the team as a side project, at least structurally.
How this echoes OpenAI’s Superalignment team
Leike’s new role resembles the mission of OpenAI’s recently dissolved Superalignment team. At OpenAI, that group had an ambitious goal: solving the core technical challenges of controlling superintelligent AI in the next four years.
According to the source article, the OpenAI team often found itself hamstrung by OpenAI’s leadership. Leike’s move to Anthropic therefore carries more than ordinary hiring significance. It shifts a researcher associated with OpenAI’s superalignment effort into a rival lab that is building its own team around related problems.
Sam Bowman, an Anthropic researcher, also commented on X, writing that Leike had led "some seminally important work on technical AI safety." Bowman said they would be leading twin teams aimed at different parts of the problem of aligning AI systems at human level and beyond.
Why Anthropic is a natural landing place
Anthropic has often attempted to position itself as more safety-focused than OpenAI. That context makes Leike’s move especially relevant in the broader AI safety debate.
The company’s own origin story is tied to OpenAI. Anthropic’s CEO, Dario Amodei, was once the VP of research at OpenAI and reportedly split with OpenAI after a disagreement over the company’s direction, specifically OpenAI’s growing commercial focus.
Amodei brought a number of ex-OpenAI employees with him to launch Anthropic. That group included OpenAI’s former policy lead Jack Clark.
Leike now becomes another major OpenAI-linked figure to join Anthropic. His arrival gives Anthropic a new focal point for superalignment research at a time when questions about AI safety, AI security and the control of advanced systems remain central to competition among leading AI labs.
The larger signal
The facts of the move are straightforward: Leike left OpenAI, criticized its approach to AI safety, and joined Anthropic to lead a new team. But the broader signal is that technical AI safety talent is becoming part of the strategic identity of major AI companies.
Anthropic is not only hiring an individual researcher. It is organizing an effort around scalable oversight, weak-to-strong generalization and automated alignment research, with existing researchers expected to move under Leike’s leadership.
For readers tracking the future of AI development, the key point is that the debate over alignment is not only public rhetoric. It is also shaping teams, reporting lines and research agendas inside the companies building advanced AI systems.