Google co-founder Sergey Brin has recently pressed key AI team members to prioritize the Gemini model as Alphabet works to narrow the lead held by competitors, according to two sources familiar with the discussions. Following Anthropic’s release of a preview for its Claude Mythos model, Brin spoke to hundreds of staff at an April gathering and called on the Google DeepMind unit to accelerate its pace, one source said. Gemini briefly led rivals last November, yet fresh releases from Anthropic and OpenAI returned it to a trailing position. In August, Google postponed the next major Gemini version by two months after tests revealed weaker results in coding tasks compared with competitors. On August 5, the company announced a major leadership change at DeepMind, with chief Demis Hassabis moving to chair and deputy Koray Kavukcuoglu taking operational control. At the same time, the two original technical leads for Gemini departed to launch a new startup. Interviews with seven people close to the project indicate the reorganization supports Alphabet’s goal of achieving dominance in AI models. The moves further reduce DeepMind’s independence, which has been limited since its 2014 acquisition as Google seeks greater commercial use of its AI work. Some teams are shifting from DeepMind into the main corporate structure. Google stock fell 4 percent on the announcement day. At a recent all-hands meeting, leaders told staff that daily operations would stay unchanged and Kavukcuoglu would continue overseeing Gemini. Hassabis outlined plans to emphasize long-term scientific and societal impacts in his new role. Staff expressed concerns that DeepMind’s independence was being further reduced. Cloud division leaders expect the changes to ease tensions over limited AI chip supplies.
