Accessibility help Skip to navigation Skip to main content Skip to footer

Opinion Business Insight

Google seeks a sharper focus in AI after Hassabis move

New role for the DeepMind co-founder comes as the competitive frontier in the technology is shifting

Add to myFT

Get instant alerts for this topic

Manage your delivery channels here Remove from myFT

Demis Hassabis speaks on stage with hands raised, in front of a vivid pink and red background at SXSW London.Demis Hassabis says he wants to spend more time on artificial general intelligence, an important threshold in AI development© TOLGA AKMEN/EPA-EFE/Shutterstock

Richard Waters

PublishedAugust 6 2026

Jump to comments section Print this page

Unlock the Editor’s Digest for free

Roula Khalaf, Editor of the FT, selects her favourite stories in this weekly newsletter.

Google has irons in more AI fires than any other big tech company. From its own specialist chips and cloud platform to its frontier models and search results, the company has carved out a broad position in AI. But this very breadth has exposed the need for sharper strategic focus.

That is one message from the shake-up in the ranks of its top AI executives this week. Demis Hassabis stepped back from day-to-day management of AI research at the same time that four top AI figures, including veteran engineer Jeff Dean, left for their own AI start-up.

One clear message: Google is still grappling with how best to marshal its considerable resources as it struggles to keep an edge in frontier AI research.

At the same time, the upheaval points to the new opportunities being thrown up by the widening boundaries in AI development. The big frontier models have been at the centre, but that is changing as the gap between the leaders narrows and stronger competition emerges from Chinese models. Instead, more of the attention is shifting to new agent-based systems with the potential to transform wide areas of research and knowledge work. For the departing Google executives, this looked like too good a moment to miss.

Google DeepMind, the AI organisation Hassabis had run and where he will now be chair, had brought together group operations in early 2023, a moment when Google was reeling from the launch of ChatGPT. This involved merging its various AI research efforts into a frontier lab with a broad remit to compete with OpenAI and Anthropic.

Three years on, this looks ready for refinement. While Hassabis had the title of CEO, his replacement will be a senior vice-president, implying less autonomy in how the unit operates and closer alignment with other Google divisions. Hassabis said he was stepping back to “focus on the big picture”, with a wider remit that includes becoming chief scientist of the parent holding company Alphabet.

This looks like part of a wider effort to refocus on Gemini (the name of both Google’s advanced frontier AI model as well as the consumer app that runs on top of it). There had already been signs that the company was reshuffling AI researchers to work on Gemini, away from other priorities. Explaining the executive moves, CEO Sundar Pichai conceded there were “areas we need to improve”.

It is little more than a year since Google first came up with a frontier model capable of holding its own against OpenAI and Anthropic, easing worries that it had fallen behind and prompting a strong rally in its share price. Now, though, Gemini is perceived to have fallen behind again and to have lost ground in the booming market for AI coding agents.

The upheaval in Google’s senior AI ranks also suggests that the capabilities of the leading AI models may be reaching an important threshold, artificial general intelligence. Hassabis said he wanted to spend more time on AGI, which he claimed was “close at hand”. While there is little agreement on exactly what this supposed breakthrough moment in AI development will entail, the leading models are already opening up significant new vistas of research.

One sign of this has been the departure of Dean and three other top Google AI figures. Their new start-up, Discovery Loop, aims to transform large areas of scientific research by automating the basic scientific process. They aren’t alone in believing that AI has reached a level where agentic AI systems like this could revolutionise scientific research. Anthropic announced a similar initiative, called Claude Science, in June. Hassabis, meanwhile, said he will also be “leaning in” more to his work leading Isomorphic Labs, a drug discovery start-up he created inside Alphabet in 2021.

With the advent of agentic systems to work in fields like this, the competitive frontier in AI is shifting. Increasingly, it resides in the software “harnesses” that control how AI agents call on and apply underlying frontier models, as well as the extra data used to boost their domain knowledge.

Coding, scientific research and generic forms of white-collar work (through Claude CoWork) are among the areas the AI agents are already toiling over. The boundaries are likely to expand quickly. Elon Musk, for instance, said this week that his frontier model, Grok, is being trained on SpaceX’s entire corpus of engineering data in an attempt to make it the best AI system for use in engineering. As Google overhauls its AI leadership, change is in the air.

richard.waters@ft.com

Reuse this content(opens in new window) CommentsJump to comments section

Follow the topics in this article

Add to myFT

Add to myFT

Add to myFT

Add to myFT

Add to myFT

Comments

Close side navigation menu

Search the FTSearch

Subscribe for full access

Read Original at Financial Times →