LONDON — Google DeepMind lost four researchers this week to rivals Anthropic and OpenAI, chief among them Nobel laureate John Jumper, as the fight for artificial intelligence talent turns cutthroat.
Jumper shared the 2024 Nobel Prize in Chemistry for AlphaFold, the DeepMind program that predicted the structures of proteins. He's now bound for Anthropic. The other three cross over to Anthropic and OpenAI as well.
The departures raise a plain question in Mountain View and London alike: can DeepMind stay at the front? Fortune reports insiders are asking it out loud. The lab that once set the pace now watches its people set up shop across town.
In this game a handful of names can move a whole field. The people who build the models are the models' edge, and a rival that lands them buys a head start money alone can't. That's why every hire reads like a headline.
Here's the rub, and it's a good one. The labs doing the poaching aren't only trading punches.
OpenAI, Google DeepMind, and Anthropic marched into the policy arena together this week, asking lawmakers for legislation to screen synthetic DNA orders. The fear runs plain: a model smart enough to help a stranger design a dangerous pathogen, and a mail-order lab willing to print it. Rivals at the hiring desk, allies at the podium.
Screening would flag suspect gene sequences before a supplier ships them. All three labs want it on the books. That's teamwork among outfits that spent the same week raiding one another's rosters.
The talent war runs a second front, too. Microsoft stood up a new division — Microsoft Frontier Company — and staffed it with 6,000 resident engineers to embed AI inside customer operations. Call it forward-deployed muscle, an army instead of a bidding war.
So the map reads plain. DeepMind ships out brains. Anthropic and OpenAI take them in, while Microsoft drills its own regiment.
Jumper's name carries freight. A Nobel medal doesn't walk out the door every week, and Anthropic just pinned one to its lapel. His AlphaFold work reshaped biology labs the world over, and now it reports to a competitor.
Whether DeepMind's remaining bench can answer is the open question. The lab still holds deep talent and Google's checkbook behind it. But momentum, like a good reporter, chases the byline — and this week the bylines moved.
One fact stands clear off the wire. The same three shops clawing for every researcher still agree on one point: nobody wants their machines teaching a stranger to brew a plague. In this business, you fight for the help and shake hands on the peril.