Demis Hassabis, the co-founder and leader of Google DeepMind, is stepping back from running the day-to-day operations of Google's main AI research lab. He will take on a new title, chief scientist for Alphabet (Google's parent company), and will keep leading Isomorphic Labs, a separate Google company that uses AI to help discover new drugs. Axios reported the changes on August 5, 2026. Koray Kavukcuoglu, who was the Chief Technology Officer at DeepMind, will be promoted to senior VP of Google DeepMind and will report directly to Google CEO Sundar Pichai. Hassabis said he made the decision to focus on the big picture and the scientific side of AI development. Axios

The changes come during a period of significant internal upheaval for DeepMind. John Jumper, a US scientist who shared the 2024 Nobel Prize with Hassabis for their work on AlphaFold, is leaving Google DeepMind for Anthropic. AlphaFold is an AI tool that predicts the shape of proteins, the building blocks of life, which had been a decades-old challenge in biology. Anthropic is another major AI company known for its focus on building AI safely. Hassabis said he believes the consolidation of talent into Anthropic will bring world-class talent together to create responsible AI breakthroughs. Reuters

Jeff Dean, who was Chief Scientist at Google DeepMind as of 2024, is also leaving. He and Google senior fellow Sanjay Ghemawat are leaving to launch Discovery Loop, an independent company structured as a public benefit corporation, meaning it is legally required to pursue a social or environmental good alongside any financial goals. Google will serve as an investor and cloud provider for the new venture. The Axios report did not specify what Discovery Loop will focus on researching. Axios

The leadership changes also follow a public resignation from within DeepMind's ranks. A Google DeepMind researcher resigned over the company's contract with the US Defense Department, publishing a resignation note exceeding 2,000 words explaining the decision. The resignation was first reported by Business Insider on July 15, 2026, and subsequently covered by the Times of India on July 17. Business Insider Times of India

Separately, Hassabis has been meeting with lawmakers in Washington to discuss a proposal for an industry-funded AI watchdog, according to CNBC. The model is similar to FINRA, a body that oversees the US financial industry on behalf of itself, but applied to AI. That effort suggests Hassabis is simultaneously stepping back from DeepMind operations and engaging on AI governance at the federal level. CNBC

Kavukcuoglu, a longtime DeepMind CTO, now takes over operations of what is arguably Google's most strategically important research organization. The fact that he reports to Pichai rather than to Hassabis signals that Alphabet leadership wants direct oversight of DeepMind's direction at a time when competition with Anthropic, OpenAI, and others is intensifying.

Jumper's move to Anthropic is notable because of the expertise involved. AlphaFold has been one of the most important applied-AI breakthroughs of the past decade, and a co-recipient of that work moving to a rival lab focused on AI safety is the kind of talent flow that could shift research priorities on both sides. Hassabis's public framing of the move as a net positive for responsible AI development is diplomatic, though whether Anthropic's research agenda will extend into the scientific domains where Jumper has worked remains an open question.

Dean and Ghemawat's departure to form Discovery Loop as a public benefit corporation, with Google as investor and cloud provider, is an interesting structural choice. That type of company builds mission obligations into its governance, which can shield research from pure commercial pressure. Google's continued financial and infrastructure involvement keeps the work within its ecosystem while granting Dean and Ghemawat the independence that a separate company provides.

The Pentagon-contract resignation adds another layer. DeepMind has historically positioned itself around beneficial AI, and internal disagreement over military applications is not unique to this organization. But a public 2,000-word resignation note forces the tension into the open. It also contextualizes Hassabis's Washington meetings on an industry-funded AI watchdog; internal tensions over defense contracts and external pressure on AI governance are running in parallel.

Taken together, the picture is one of an organization in substantial transition. The person who has been the public face of Google's AI ambition is deliberately stepping back from operations to concentrate on scientific direction at the Alphabet level, while his CTO takes the operational helm, a Nobel laureate departs for a competitor, two senior engineers spin out a new venture, and a researcher exits over an ethical line. That is a dense concentration of change for any single organization, and it arrives while the AI industry's competitive and regulatory dynamics are themselves in rapid motion.

In the most generous reading, this restructuring enables a division of labor that lets Hassabis operate at the altitude where Alphabet-level scientific strategy actually gets set, while Kavukcuoglu handles the day-to-day execution. Whether that division holds under competitive pressure from well-funded rivals and internal talent churn is the question the rest of 2026 will answer.