Leadership changed at Google DeepMind: Hassabis becomes chair, Jeff Dean leaves after 27 years
According to Sundar Pichai's post on the company blog, Koray Kavukcuoglu takes over day to day leadership of Google DeepMind. Demis Hassabis steps out of operations to become chair of Google DeepMind and chief scientist of Alphabet. Jeff Dean is leaving after 27 years and founding an independent public benefit corporation with Sanjay Ghemawat.
Three names and three new roles in a single post
Google reorganized the senior leadership of its AI work. Sundar Pichai's post on the company blog carries three headlines. Koray Kavukcuoglu becomes senior vice president of Google DeepMind, taking over Gemini model development, frontier AI research and the related teams, reporting directly to Pichai. Demis Hassabis steps back from day to day leadership to become chair of Google DeepMind and chief scientist of Alphabet, shifting his focus to long range AI strategy. Jeff Dean is leaving the company.
Per Search Engine Journal's August 5, 2026 report, Dean is leaving after 27 years at Google to launch an independent public benefit corporation named Discovery Loop together with Sanjay Ghemawat. Its stated purpose is accelerating discoveries in machine learning, science and engineering.
Pichai frames the change as forward looking, describing the company as being at a dynamic moment: "We are at a dynamic moment with so much opportunity ahead."
Who picks up what
Per Search Engine Journal, before this appointment Kavukcuoglu served as chief technology officer of Google DeepMind and chief AI architect at Google, and he keeps the chief AI architect title alongside the new role. His day to day remit covers the Gemini models and the frontier model teams.
On the Hassabis side the change is more than a title. The same report describes his new position as a broader remit across Alphabet's AI divisions, Isomorphic Labs included. So the DeepMind co founder is not stepping away from the company, he is moving toward long range research and strategy.
Jeff Dean's exit carries the heaviest symbolic weight. Search Engine Land's August 6, 2026 report recalls that Dean played a defining role in building early Google Search and later AI layers such as RankBrain and AI Overviews. The same report says Oriol Vinyals and Quoc Le are joining the new organization alongside Ghemawat.
Nothing in the announcement changes a product. There is no timeline for the Gemini roadmap, model releases or the AI surfaces in Search. The sources carry no information on that, so for today the only thing that changed is who runs what.
The line between research and product is redrawn
It is worth not overstating the business impact here: nothing changes in anyone's ad account or search results tomorrow morning. Over the medium term, though, two signals can be read.
First, putting day to day control of Gemini and the frontier models in one person who reports directly to the CEO points to an intent to shorten the distance between model development and product. Kavukcuoglu keeping the chief AI architect title supports that reading. This assessment is ours.
Second, Hassabis moving toward long range research while Dean builds something new outside the company continues a pattern the industry has seen for a while: senior researchers stepping out of large company structures into focused organizations. This does not mean Google's AI capacity is shrinking, and the sources make no such claim. But where the names gather is a hint about which centers will shape the coming years.
The third point is more practical. In many markets, Türkiye included, business visibility now depends not only on classic search results but also on surfaces such as Gemini and AI summaries. A change at the top of the team running those surfaces can be read as a signal of how much corporate priority that area carries. That reading is ours too.
What it means for businesses in Türkiye
The assessment below does not appear in the sources, it is our reading. The sources carry no Türkiye specific breakdown or local note.
A leadership announcement changes none of the work a business in Türkiye has to do. What it changes is the planning horizon. In a period where Google keeps moving AI further toward the center, measuring a company's digital presence purely by asking "where do I rank" is increasingly incomplete.
There are three concrete places to look. First, whether the content is legible to machines: clear headings, paragraphs that answer questions, properly structured data. Second, the consistency of the corporate identity: business name, address, service description and contact details being identical everywhere is a precondition for an assistant to describe you correctly. Third, measurement: being able to separate traffic from these surfaces inside analytics.
None of this started with this news, and none of it ends with it. It can be noted as a sign that the agenda keeps flowing in this direction. That last sentence is our comment.
The UNALSOFT take
In our agentic AI work we make a principle of not binding a business to a single model or a single provider. This news is a reminder of why. Teams, priorities and roadmaps on the provider side can change. If a company's AI layer has to be rebuilt whenever the model behind it changes, that layer was designed fragile from the start. In the systems we build the value does not sit in the model, it sits in the company's own data, its process definition and the flow that carries that process. Models change, the system stays.
Is your AI layer resilient to model changes?
Let's review your current setup together and map the dependencies.