Lilian Weng is an AI researcher known for her work on safety systems and for a long-running technical blog on alignment and machine learning. She led the Safety Systems team at OpenAI, left to co-found Thinking Machines Lab in 2025, and returned to OpenAI in 2026 to work on recursive self-improvement.
Roles
At OpenAI, Weng led Safety Systems, the function responsible for practical deployment safeguards on shipped models. She was among the group of senior OpenAI research and post-training staff — with John Schulman, Barret Zoph, Jonathan Lachman and Andrew Tulloch — who founded Thinking Machines Lab under Mira Murati in 2025. The concentration of that group's safety and post-training experience shaped external expectations of the new company's safety posture (Thinking Machines Lab).
Her return to OpenAI became public on August 3, 2026, days after she said that being a co-founder was taking a toll on her health and that she wanted a more focused position. She is the fourth Thinking Machines co-founder to leave the company within the year to that date. The reported remit on return is recursive self-improvement (Source: axios.com). See Recursive Self-Improvement (RSI).
Context of the move
The departure sits within a wider pattern of researcher movement among frontier labs during 2026. In the same reporting, a source familiar said Dario Amodei has expressed concern that new hires are coming for the money rather than the mission, and an executive technology recruiter attributed movement partly to compensation and "some ego about changing the world"; more than 400 former Apple employees were reported to be working at OpenAI (Source: axios.com). These are attributed characterizations of the labor market rather than statements about Weng's own reasons, which she gave as health and focus.
Writing
Weng is a widely-read technical writer on alignment, agents, and hallucination, publishing long-form explanatory posts that are cited as reference material within the field.
Relationships
- related: OpenAI — employer, before and after Thinking Machines
- related: Thinking Machines Lab — co-founder, 2025 to 2026
- related: Mira Murati — Thinking Machines founder and CEO
- related: Recursive Self-Improvement (RSI) — stated focus on returning to OpenAI
- related: Dario Amodei — quoted in the same reporting on frontier-lab hiring motives