Google DeepMind Shifts Daily Control as Its EMEA Hiring Share Falls to 18.6%

Google DeepMind’s daily operations are now in Koray Kavukcuoglu’s hands, while Demis Hassabis moves into chairman and Alphabet chief scientist roles. The handoff comes as DeepMind’s share of research and advanced-engineering hires in Europe, the Middle East and Africa fell to 18.6%, from 49% three years earlier, according to Zeki Data.

A new split between strategy and execution

The restructuring separates operational control of the lab from Hassabis’s wider remit. Hassabis remains in charge of Isomorphic Labs, Alphabet’s drug-discovery business, and is expected to focus on strategic and global AGI matters while advising DeepMind. Google framed the change as part of staying at the frontier while shaping the future of artificial general intelligence.

The leadership map

• Hassabis becomes DeepMind chairman and Alphabet chief scientist, while continuing to lead Isomorphic Labs.
• Kavukcuoglu, formerly DeepMind’s CTO, takes responsibility for day-to-day AI development and frontier-model work.
• Jeff Dean, a 27-year Google veteran, has left to co-found Discovery Loop, which aims to automate scientific research. Sanjay Ghemawat, Oriol Vinyals and Quoc Le are joining him.

The recruitment buffer has narrowed

Zeki’s figures put the leadership decision in a more difficult competitive setting. DeepMind is still adding more research and advanced-engineering staff globally than it loses, but its arrivals-to-departures ratio dropped from roughly 12 hires for every exit in the second quarter of 2023 to about two for every exit in the third quarter of 2026. That is a smaller cushion for a lab whose progress depends on a limited pool of specialized technical staff.

49%EMEA share of specialized hires

Zeki Data estimates DeepMind’s share of research and advanced-engineering hires across Europe, the Middle East and Africa declined from 49% in 2022–23 to 18.6% in 2025–26.

12 to 1Arrivals for each departure

Zeki Data estimates DeepMind’s arrivals-to-departures ratio declined from about 12-to-1 in the second quarter of 2023 to roughly 2-to-1 in the third quarter of 2026.

The regional-share figure is a measure of the portion of hires, not a direct count of DeepMind’s total workforce or a measure of model quality. Zeki tracked 20,900 people in research and advanced-engineering roles at 10 companies using public information, so its dataset may undercount the overall population. Within that framework, it found annual research-and-engineering headcount growth of about 27% at DeepMind, versus 97% at OpenAI and 152% at Anthropic; the latter two started from smaller bases.

Losses concentrate in contested research areas

The talent movement is not evenly distributed. Of people who left DeepMind during the preceding 12 months, Zeki found that 25% went to Anthropic, 21% to Meta and 14% to OpenAI. It also identified a net loss in large language models and multimodal systems: 19.2% of leavers specialized in those fields, compared with 15.6% of joiners.

The lab has also seen departures from a signature scientific program. Thirteen of the 29 named authors on the AlphaFold2 paper have left DeepMind since publication, according to Zeki. Fortune’s interviews with current and former DeepMind staff pointed to aggressive recruiting by rivals, morale concerns, and a growing emphasis on Gemini products as factors in the lab’s changing appeal.

Research openness remains part of the retention question

DeepMind’s publication policy has become a point of friction in that transition. The Financial Times reported in 2025 that the lab introduced a tougher internal review process and a six-month embargo for some strategically sensitive generative-AI papers. DeepMind said at the time that it remained committed to publishing research and was updating policies to preserve teams’ contributions to the wider research ecosystem.

Kavukcuoglu inherits the immediate job of directing DeepMind’s model work while Hassabis takes on a broader scientific and strategic role. The unresolved test is whether that division can improve recruitment and retention in fields where Zeki measured a net loss in large language models and multimodal systems. The leadership change itself does not answer that question.