DeepMind Leadership Transition Report, Conflicting Official Titles

Google DeepMind is reportedly transitioning from a research leadership centered around co-founder Demis Hassabis to an operational system led by Chief Technology Officer Koray Kavukcuoglu. While the official title arrangements are still conflicting, the delay in the Gemini launch and the movement of key personnel have highlighted a shift in the organization's focus.

Fortune reported that on the 13th of the month at 8:13 PM KST, Hassabis stepped back from day-to-day operations, and Kavukcuoglu took over the practical responsibilities. The same report also mentioned the departure of Google Senior Scientist Jeff Dean.

However, Google DeepMind's official introduction page still lists Hassabis as co-founder and CEO. Kavukcuoglu is also described on Google's blog as the Senior Vice President of Google DeepMind and Google's Chief AI Architect. This indicates a gap between the latest reports and the official profiles regarding titles.

The core of this reorganization is the change in DeepMind's role rather than just personnel movement. DeepMind started as a London-based research organization but has become more closely linked with products like Gemini, Gemini Robotics, and Lyria since merging with Google Brain in 2023.

The Guardian interpreted this change as a trend of the London research organization moving deeper into Mountain View's Google product strategy. It explains that DeepMind is maintaining its independent research institute character while increasing connections with commercial products like Google Search, Cloud, and Model API.

Previously, we reported on the change in the day-to-day operational leadership of Google DeepMind and the restructuring of the AI research organization. At that time, the report indicated that Hassabis would focus on long-term research strategy while Kavukcuoglu would oversee Gemini development and daily operations in a division of labor structure.

Product competition has also been a burden. Axios reported in July that Google DeepMind's Gemini 3.5 Pro was delayed by several months, and employee morale decline affected the launch delay. Google denied interpretations that morale decline led to poor model performance and claims of large-scale personnel departures.

Gemini is Google's generative AI model product line. Developing frontier AI models requires not only research outcomes but also learning computing resources, product application, and safety reviews, which means coordination between research and product organizations can impact launch schedules.

Personnel movements have continued. Former Google DeepMind researcher John Jumper announced in June that he would move to Anthropic, and Fortune reported that Noam Shazeer also moved to OpenAI. Business Insider reported that Dean established the Discovery Loop with Sanjay Ghemawat, Oriol Vinyals, and Quoc Le.

In Dean's case, there is also a discrepancy between his official profile and the latest reports. Google Research and Google's blog still list him as a Chief Scientist. The personnel movement reports came out before any updates to the official pages.

The development flow of Google DeepMind has not stopped. The official model card list was updated as of July 21 to include Gemini 3.6 Flash and Gemini 3.5 Flash Lite. DeepMind also separately introduced the Gemini 3.5 Flash Cyber model for security purposes.

Hassabis's interests appear to be shifting from operations to AI safety and long-term research. In July, he unveiled plans for an independent standards body to pre-assess frontier AI models. TechCrunch and Axios reported that Sam Altman and Elon Musk responded positively to this.

Market reactions have shifted towards viewing organizational stability and AI investment costs together. The Financial Times reported that Alphabet's stock fell by 5% following this reorganization. The article mentioned that organizational stability and AI investment costs were concerns for investors.

In the Korean market, this matter serves as a reference case for observing big tech AI investments and model competition. We previously reported that Alphabet emphasizes a structure that connects Gemini with model APIs, cloud, TPU, and server infrastructure to expand AI usage. The adjustment of DeepMind's role aligns with the trend of linking research organization performance to product sales and infrastructure investment recovery logic.

The debate over whether Google DeepMind is a research institute or a product engine will be reaffirmed in the official title arrangements and the next Gemini model release. Currently, there is a discrepancy in the title listings of Hassabis and Kavukcuoglu between the confirmed official pages and reports.

Verification sources: Fortune, Google DeepMind, Google Blog, Axios, The Guardian, Business Insider