33 researcher moves between frontier labs since 2025: Meta pulled hardest, then Thinking Machines gave back
Asked:
“Which AI researchers moved between frontier labs since 2025?”
From January 2025 through 11 November 2026, public reporting recorded 33 direct moves by 29 named technical AI researchers between frontier-model labs — OpenAI, Google DeepMind, Anthropic, Meta's superintelligence effort, Thinking Machines Lab and Safe Superintelligence. This is a public-reporting census, not a complete employee registry; every move was corroborated across at least two independent hosts.
Flow of reported moves between labs, ribbon width = number of researchers. Hover a ribbon for the names; hover a lab for its ins and outs. Google = Google DeepMind; Meta = Meta Superintelligence Labs / FAIR.
The busiest corridor was OpenAI → Meta: 10 researchers, 8 of them in June 2025 alone, including Shengjia Zhao as chief scientist of the Superintelligence Lab. reuters.com
Thinking Machines Lab hired 8 and lost 5: four founding-team members — Barrett Zoph, Lilian Weng, Luke Metz and Sam Schoenholz — returned to OpenAI in January 2026. techcrunch.com
Google DeepMind lost 4 researchers to Anthropic in June 2026, including AlphaFold Nobel laureate John M. Jumper. giftarticle.ft.com
Andrew Tulloch moved twice — OpenAI to Thinking Machines (February 2025), then on to Meta (October 2025) — the only person to touch three labs. techcrunch.com
Moves per month
All 33 reported moves
Month
Person
From
To
Role at destination
Source
Public-reporting census of direct researcher moves between frontier-model labs, 1 Jan 2025 – 11 Nov 2026. 33 rows, one per person-route-month; 29 named people, 4 repeat movers. Month is the first reliably reported month, which may differ from the actual start date. Each move corroborated on at least two independent hosts; the most-used single URL backs 7 of 33 rows. Product, policy, operations and non-frontier moves excluded.