Meta is losing one of the highest-profile researchers it recruited during last year’s artificial intelligence talent war.

Andrew Tulloch is leaving Meta to join Anthropic, according to The Wall Street Journal. The move comes less than a year after Tulloch returned to Meta as the company was spending aggressively to assemble its superintelligence team.

Tulloch has moved through several of the most important organizations in frontier AI. He spent roughly 11 years at Meta before leaving in 2023 for OpenAI, later became a co-founder of Thinking Machines Lab, and then returned to Meta in 2025.

His recruitment became one of the most widely discussed examples of the AI industry’s escalating compensation race. Reports at the time described a potential Meta package worth more than $1 billion and, under some assumptions, as much as $1.5 billion over several years. Meta disputed the higher-end characterization, calling reports of a $1.5 billion offer inaccurate.

AI talent is behaving like strategic infrastructure

Tulloch’s departure is notable not simply because of the reported compensation. Frontier labs increasingly treat a small group of researchers as strategic assets capable of materially influencing model quality, research direction and the speed at which new systems reach market.

That has created an unusual labor market in which researchers can move among companies carrying valuations in the hundreds of billions of dollars — and where losing a handful of people can become a competitive signal.

Anthropic, meanwhile, continues to recruit as it competes directly with Meta, OpenAI and Google for both enterprise customers and frontier-model leadership.

The move also arrives just after Meta launched Muse, its new personal AI agent, underscoring the contrast between Meta’s expanding product ambitions and the continuing churn inside the research organizations responsible for delivering them.

For Meta, the question is whether its enormous spending can create a research organization with staying power. For the broader industry, Tulloch’s exit is another reminder that the AI race is being fought not only in chips and data centers, but person by person.