CALIFORNIA: Two of Google’s most prized artificial intelligence researchers are walking out the door, and straight into the arms of Anthropic. Jonas Adler and Alexander Pritzel, both viewed internally as key contributors to Google’s Gemini model, are set to move to the Claude maker, according to people familiar with the matter who spoke on condition of anonymity because the information is not public. It is the latest in a string of high-profile departures threatening to undercut the search giant’s standing in the AI race.
Adler worked on Google’s AI coding effort, while Pritzel was involved in training the company’s AI systems. Google, an early pioneer in artificial intelligence, spent much of the current boom playing catch-up with OpenAI and Anthropic before finally hitting its stride late last year with sharper models and better chips. The timing of this latest blow stings: the company had already lost two prominent staffers in recent days, with Nobel laureate John Jumper heading to Anthropic and star researcher Noam Shazeer defecting to OpenAI.
The market noticed. Alphabet shares closed down slightly after falling as much as 1.2 per cent during Wednesday’s trading session, as investors weighed fresh doubt over Google’s ability to keep pace in the fiercest model-building contest the tech industry has seen.
The exits underline the pressure Google faces from two rivals on the cusp of going public, both dangling the kind of rare pre-IPO payday that can tempt even well-compensated Big Tech staff to jump ship. In at least one case, the departure appears to have been preceded by friction over computing resources, an issue that has already pushed other employees out the door entirely. Shortly before Shazeer announced his move to OpenAI, computing power earmarked for one of his projects was reassigned to a London-based team at Google DeepMind, according to two people familiar with the matter, ostensibly to sharpen collaboration and streamline the company’s pre-training work. Adler, Pritzel, Jumper and Shazeer did not respond to requests for comment. Anthropic declined to comment. A Google spokesperson said the company remains confident in its position in the market for AI talent, pointing to recent remarks from DeepMind chief executive Demis Hassabis.
Hassabis, speaking at an event in Cannes, did not dodge the scale of the problem. There is a lot of talent movement between all the leading labs and Google wins its fair share of the top talent, he said, while conceding the market for AI talent is the most ferociously competitive the tech industry has ever seen.
Shazeer’s own career arc tells the story of just how brutal that market has become. He co-authored a paper that helped catalyse the entire AI boom, left Google in 2021 to found the chatbot startup Character.AI, then rejoined the company in 2024 as part of a licensing deal that valued his outfit at $2.5 billion. Once back, he co-led development of Gemini and was reportedly working on a new AI architecture, still built on the transformer technique he helped invent in 2017, that was already producing promising results before he walked.
Shazeer was as divisive as he was admired inside Google, with comments on transgender identity and the Gaza conflict stirring controversy among staff. Jumper, by contrast, became a face of the company’s most ambitious AI push after winning the Nobel prize for landmark work using AI to predict protein folding. Adler and Pritzel, both now bound for Anthropic, worked with him on that very research, and key members of Jumper’s protein-folding team have been leaving Google DeepMind for months, some heading to Isomorphic Labs, the Alphabet spinout chasing AI-designed drugs.
The numbers make grim reading for Google. DeepMind engineers are now nearly eleven times more likely to leave for Anthropic than the other way round, according to a 2025 analysis from venture capital firm SignalFire. Anthropic, which both competes with Google and partners with it, has aggressively siphoned talent from the search giant while chasing its own life-sciences and healthcare ambitions. The Claude maker recently raised a fresh funding round at a $965 billion valuation, overtaking OpenAI, and is reportedly weighing a public listing as soon as this autumn.
British non-compete law offers Google a brief reprieve. UK-based AI researchers, where DeepMind’s leadership sits, are typically bound by lengthy non-compete agreements, meaning Jumper likely will not start at Anthropic until next year. It is cold comfort. Another departing researcher, Arthur Conmy, announced on X that he too was headed to Anthropic to work on AI safety, having spent his DeepMind tenure contributing to the Gemini 2.5 model and AI coding work.
Google insists it still has the deepest bench in the business. The exit queue at Mountain View tells a rather different story, and with Anthropic eyeing a stock market debut and a near billion-dollar valuation to dangle, the talent war shows no sign of calling a truce.