Google Reshapes AI Leadership as Hassabis Becomes Alphabet Chief Scientist
TL;DR
Demis Hassabis will become chair of Google DeepMind and Alphabet chief scientist, while Koray Kavukcuoglu takes operational responsibility for models, research and Gemini product teams.
The operational test for this leadership change will arrive within three to six months. If Google releases its planned new models on schedule while the Gemini app continues to add active users, separating product execution from broader scientific strategy may have produced a measurable benefit. Delayed releases, stagnant adoption, or less clear decision rights would undermine that explanation. The announcement defines reporting lines, but it does not state when every appointment formally takes effect or publish internal performance targets.
Google chief executive Sundar Pichai announced the Google DeepMind reorganization on August 5, 2026. Demis Hassabis will give up day-to-day operating responsibility and become chair of Google DeepMind and chief scientist of Alphabet. He will continue to lead Isomorphic Labs and advise on models, research, and global questions about artificial general intelligence. Reuters independently reported the transition and confirmed that operational management moves to Koray Kavukcuoglu, currently Google DeepMind’s chief technology officer and Google’s chief AI architect.
One reporting line for models, research, and products
Kavukcuoglu will serve as senior vice president of Google DeepMind and report directly to Pichai. His remit includes Gemini model development, frontier AI research, the Gemini app, and developer teams. He has spent 13 years at DeepMind, where Google says he established the deep-learning team and helped lead work behind WaveNet and DQN. The structure places research, model engineering, and product adoption under one operating leader. Hassabis, by contrast, will spend more time on AGI strategy, scientific questions, and the drug-discovery work of Isomorphic Labs.
Google framed the change against the scale of its current products. The company says the Gemini app has 950M+ monthly active users and that Gemma models have passed 900M+ downloads. It also says demand for Flash is strong, its Cyber model is live, and Hassabis’s employee note referred to work on Gemini 4. These figures all come from Google. Reuters verifies the personnel changes, but its report does not independently audit Gemini usage, model quality, research spending, or the causal claim that a new structure will accelerate releases.
Dean and Ghemawat leave to form a company
Jeff Dean, after 27 years at Google, will leave with Google Senior Fellow Sanjay Ghemawat to create an independent public-benefit corporation focused on discoveries in machine learning, science, and engineering. Google says it will participate as a founding investor and cloud partner. The two organizations also plan to collaborate on a research framework for machine-learning systems and related infrastructure advances.
Important commercial terms remain undisclosed. Google did not name the new company, specify its investment, state an ownership percentage, or provide a headcount. The announcement therefore does not show how much technical knowledge the partnership retains inside Google’s orbit, whether employees will follow Dean and Ghemawat, or whether the new company will eventually compete with Google research programs. Those uncertainties matter because Dean and Ghemawat helped build infrastructure that supported both Google’s early search systems and later neural-network work.
The reorganization now assigns three distinct responsibilities. Kavukcuoglu runs model and product delivery; Hassabis handles Alphabet-level science and AGI strategy; Dean and Ghemawat pursue research through a separate organization. The next useful evidence will be the release timing of Gemini 4, whether the Gemini app moves above 950M+ monthly active users, the rate of Gemma downloads, and the new company’s disclosed research output and financing terms. Until those data arrive, the reporting lines have clearly changed, while any improvement in research productivity remains unproven.
Sources:
Related Articles
Google Launches Gemini 3.7 Flash: Half-Price Until Year-End, With Agent Costs Still Tied to Retry Rates
Google positions Gemini 3.7 Flash as a workhorse for coding and AI agents, with higher vendor benchmarks and temporary half-price access, while architecture, training methods, and production retry rates remain undisclosed.
Google Puts SL2T on Pixel 11: Sign-Language Input Meets Real-World Error Conditions
Google DeepMind is shipping its SL2T American Sign Language translation model in Gboard and Live Transcribe on Pixel 11, while independent real-world error rates remain unavailable.