OpenAI-to-Meta accounts for more than a third of identified frontier-lab moves

A scan of reporting since January 2025 identified 38 direct moves by technical AI researchers and research leaders across frontier-model labs. The largest corridor was OpenAI to Meta, with 14 people; the next largest was Thinking Machines Lab to OpenAI, with five.

1. Meta’s 2025 recruiting wave dominates. Fourteen identified researchers went from OpenAI to Meta, and another five arrived from Google DeepMind.
2. Thinking Machines became a two-way talent hub. Two researchers went to Meta and five went or returned to OpenAI.
3. Anthropic’s inbound pattern changed in 2026. The identified moves include three Google DeepMind researchers and two OpenAI researchers joining Anthropic.
4. The list is evidence-based, not a complete personnel census. It includes direct moves with a named destination in accessible reporting and excludes departures with no confirmed next lab.

Named moves and sources

Method: web search across 1,469 initial results plus focused 2025 and 2026 searches, accessed 20 Aug 2026. Dates are the reported move month or, where only an announcement was available, the announcement month. “Frontier lab” here means an organization developing frontier foundation models; roles were limited to researchers, research engineers, and technical research leaders. One person can appear more than once only for distinct direct moves.