The AI talent revolving door: Meta's 2025 raid, then DeepMind's 2026 exodus

29 researcher moves between eight frontier AI labs — OpenAI, Anthropic, Google DeepMind, Meta Superintelligence Labs, xAI, Safe Superintelligence, Thinking Machines Lab and Tencent — traced from public reporting, January 2025 through 29 August 2026. The month shown is the first credible public report of each move.

Method: 29 curated moves from a broad public-web scan (three result sets, 49 raw rows) covering 2025-01-01 to 2026-08-29. "Month" is the move or first-report month as published, usually announcement rather than payroll start; eight June 2025 Meta recruits with no stated month are dated to the 30 June first report. Duplicate rows were merged keeping the highest source count; Google was normalised to Google DeepMind and Meta variants to Meta Superintelligence Labs. Corrections applied: Barret Zoph's Thinking Machines → OpenAI move dated January 2026; Andrew Tulloch dated October 2025; Joshua Gross dated April 2026; Zoph's OpenAI → DeepMind return dated August 2026. Rows for Andrej Karpathy and Jolene Parish were excluded as unsupported; moves with an endpoint outside the eight in-scope labs (Alexandr Wang, Johan Schalkwyk) were cut. Source count is a confidence cue only; this list is not exhaustive — many employment changes are never publicly reported.