Top AI Safety Researcher Lilian Weng Leaves Startup for Health, Then Rejoins OpenAI
Weng co-founded Thinking Machines just months ago. Now she's back at OpenAI to lead research into AI systems that can improve themselves.
Key points
- Lilian Weng stepped down as co-founder of Thinking Machines on 27 July 2026, citing health problems caused by startup stress.
- OpenAI confirmed on 30 July 2026 that Weng would rejoin the company, where she previously served as VP of AI Safety Research.
- At OpenAI, Weng will lead a team focused on recursive self-improvement, a process where an AI system rewrites and upgrades its own code to become more capable.
- Thinking Machines was founded by former OpenAI CTO Mira Murati, who publicly supported Weng's decision to prioritise her health.
Lilian Weng built her name at OpenAI as the vice president of AI safety research, the part of the company trying to ensure powerful AI doesn't cause harm. She left to help start Thinking Machines, a new lab co-founded by Mira Murati, the former chief technology officer at OpenAI. Now, weeks into that venture, Weng is heading back.
She shared the news herself. In an internal message posted to Slack, then republished on X, Weng wrote: "After thinking about it for several months, I ultimately have to admit that the amount of consistent stress and workload have pushed me beyond what my health can sustain physically."
It's a candid admission from someone at the very top of one of AI's most watched new companies. We first covered Thinking Machines on 15 July 2026 when its debut model, Inkling, launched for businesses wanting to own their AI rather than rent it.
Why is she returning to OpenAI if burnout drove her away?
Startup life is a different beast. At a company she co-founded, every problem is her problem. Back at OpenAI, where she is one senior researcher among thousands, the weight is distributed differently.
That distinction matters. First reported by TechCrunch AI, OpenAI confirmed Weng will head a new top-level team working on recursive self-improvement, building AI systems that can analyse and rewrite their own design to get better over time. It's one of the most ambitious areas in AI research, and dedicating a whole team to it signals how seriously OpenAI takes the idea.
What did her Thinking Machines colleagues say?
Murati responded with warmth. "We'll miss you, it's been wonderful building Thinky together. I'm glad that you're putting your health first. Thank you for everything," she wrote under Weng's post on X.
Whether Murati knew Weng was heading back to OpenAI specifically is not clear.
Does this matter beyond one person's career change?
Yes, for two reasons. Top AI researchers are the scarcest resource in the industry right now. Labs are competing fiercely for a small pool of people who understand how to build and govern the most powerful systems. Losing a co-founder this early is a real blow for Thinking Machines.
Weng's new role also puts her at the centre of a thorny question the whole field is wrestling with: whether an AI can eventually improve itself faster than humans can oversee it. Her safety background makes her an interesting, and arguably reassuring, choice to lead that work. Our story on 1,100-plus AI employees urging the US government to slow the race showed just how raw that anxiety already is across the industry.
For ordinary people watching AI develop from the outside, this is a reminder that the humans building these systems aren't machines. Burnout shapes who stays, who goes, and ultimately what gets built. That's worth paying attention to.
