Google DeepMind Leadership Shift Prioritizes Gemini 4 Release Over AGI Goals
AI

Google DeepMind Leadership Shift Prioritizes Gemini 4 Release Over AGI Goals

TechNews Editorial
TechNews EditorialSep 24, 2026 · 1 min read
Share

Google DeepMind is prioritizing commercial product delivery over its traditional pursuit of artificial general intelligence. Koray Kavukcuoglu made his first media appearance as head of the unit at The Information's AI Agenda Live Summit. He stated that the conversation about achieving AGI is not the right conversation.

Kavukcuoglu took over in August after Demis Hassabis stepped down as DeepMind CEO. Hassabis now chairs the unit and serves as Alphabet's chief scientist. Kavukcuoglu stated that his team is working to release Gemini 4 much earlier than the end of the year. The model is currently in an early phase of post-training.

Engineers are already using the model internally for a coding tool named Antigravity. Kavukcuoglu noted that fast iterations will follow the initial release. Google needs to make up ground after Gemini 3.5 Pro quietly vanished following missed deadlines reported by the Financial Times.

Rivals have advanced significantly during this period. Anthropic released and updated its Mythos and Opus families, while OpenAI has rolled out GPT-6 models since early September. Kavukcuoglu expressed total confidence in his team and stated that Google will remain at the technological frontier.

The restructuring follows years of changes triggered by the launch of ChatGPT. Google merged Google Brain and DeepMind into Google DeepMind, and Gemini development moved to the Bay Area. A former Google executive told the Guardian that the era of DeepMind as an independent actor is over.

The lab has also faced a notable talent drain. Departures include Noam Shazeer to OpenAI, John Jumper to Anthropic, and Jeff Dean launching a startup with other researchers. DeepMind's ratio of new hires to departures fell from twelve to one in early 2023 to roughly two to one in the third quarter of 2026.

Kavukcuoglu emphasized building trustworthy intelligent agents rather than debating AGI definitions. He noted that new models also assist hardware engineers in designing future generations of Google tensor processing units. Google plans to continue safety testing while pushing for rapid model releases.

Newsletter

Get the best AI & tech news daily

A concise daily digest. Unsubscribe anytime.

We use your email only to send this newsletter.

Related Stories