August 5, 2026, (Inside AI) — Alphabet has reshuffled its AI leadership, removing Demis Hassabis from day-to-day management of Google DeepMind and seeing the departure of four top Gemini model leaders, including Jeff Dean, to launch a new startup.
The moves come as Google's flagship Gemini 4 model remains unreleased past its planned June launch, fueling concerns that the company is losing ground to Anthropic and OpenAI. Hassabis will become Alphabet's chief scientist and chairman of Google DeepMind, while CTO Koray Kavukcuoglu takes over as senior vice president of the unit.
Dean, Sanjay Ghemawat, Oriol Vinyals, and Quoc Le have left to found Discovery Loop, a public benefit corporation researching machine learning breakthroughs. The startup has secured investment from Google and a cloud computing partnership with its cloud division.
In a memo to staff, Hassabis cited the proximity of artificial general intelligence as the reason for his shift. Alphabet CEO Sundar Pichai said in a separate memo that the new role lets Hassabis "put his full attention on actively shaping the future of AGI."
The departures extend a summer brain drain. Noam Shazeer, a Gemini co-lead, left for OpenAI in June, and Nobel laureate John Jumper joined Anthropic the same week. These exits follow Google's $2.7 billion deal in 2024 to bring Shazeer and his team back to the company.
Hassabis will also spend more time at Isomorphic Labs, his drug discovery spinout. "I've always believed the No.1 application of AI should be to improve human health," he said. "It's time for AI to prove its unequivocal value to the world, and what better way to demonstrate that than to help finally cure diseases like cancer."
The leadership overhaul marks a strategic pivot. Hassabis, long focused on research over profits, will now explore AGI's societal impacts with few direct reports. Kavukcuoglu, already Alphabet's chief AI architect, steps into a role without the CEO title, signaling a tighter integration with the parent company.
Google's AI efforts have faced mounting pressure. Gemini 3, released last November, was seen as closing the gap with rivals, but subsequent advances from Anthropic and OpenAI, along with cheaper Chinese open-source models, have put Google on the defensive. At its May developer conference, executives emphasized Gemini's cost efficiency rather than cutting-edge performance.
The Discovery Loop team brings deep expertise. Vinyals and Le coauthored a seminal 2014 paper with Ilya Sutskever that showed scaling data for pre-training could dramatically improve AI, a concept foundational to modern systems. Dean and Ghemawat were architects of Google's core infrastructure.
The shakeup raises questions about Gemini 4's delays. A Google spokesperson declined to detail the reasons for the timing, but industry observers note that model development setbacks can cascade from leadership gaps and talent flight. The company's Gemini technical report from late 2023 highlighted ambitious multimodal goals that may have proven difficult to execute.
Hassabis's memo touted "great progress" on Gemini 4, but without a launch date, skepticism persists. Meanwhile, Discovery Loop's public benefit structure suggests a focus on foundational research rather than immediate commercialization, potentially pulling critical talent away from product-driven timelines.
The reorganization echoes past Google AI restructurings. The 2023 merger of DeepMind and Google Brain was meant to unify efforts after ChatGPT's launch blindsided the company. Now, with Hassabis stepping back and core Gemini leaders exiting, the integration strategy appears to be fraying just as competition intensifies.
