Do we like them?

Do we like Demis Hassabis?

Our score: 7 out of 10. A clean record on people and the planet, plus real scientific good, held back by one contested call on military AI and a thin record on LGBTQ+ issues.

Demis Hassabis was a chess kid. He started at four, and by 13 he was a master playing adults in international tournaments. He has said chess got him “thinking about thinking itself”, and a programmable chess computer and an Amiga 500 in his early teens got him hooked on AI.

That obsession became DeepMind, which he co-founded in 2010 and Google bought in 2014 for more than US$500 million. It produced AlphaZero, which taught itself chess from the rules, and AlphaFold2, which predicted the structures of almost all 200 million known proteins. In 2024 that work won him a share of the Nobel Prize in Chemistry.

Now the harm check, and the honest bit. We found no campaigning against LGBTQ+ people, women, migrants or the climate, and no slide into conspiracy theories. But we also found nothing from him on trans rights either way, so that silence is a gap, not a pass. The contested call is military: Google updated its AI principles in 2025, and he defended it, saying “Nothing's changed about our principles”.

Where you land on that depends on whether you read it as a pragmatic response to a messier world or as a walk-back of an old promise. He says any deployment has to clear a high bar, with benefits that “substantially outweigh the risk of harm”, and he points to cyber defence and biosecurity as areas where he sees a duty to help. Critics say the change opens the door to weapons work.

On AI itself he sounds like the cautious one in the room. He says the research on whether advanced systems could slip out of human control is unresolved, and he argues for international standards and cooperation. He describes himself as a “scientist first and foremost”, which fits a career pointed at protein structures and drug discovery.

The recent shake-up is corporate rather than personal. He swaps day-to-day management for AGI strategy and science, and more time at Isomorphic Labs, amid reported delays to a Gemini model and several high-profile departures. Nothing in the coverage we read suggests wrongdoing.