Google DeepMind CEO Hassabis: Humanity Has 'Few Years' to Prepare for AGI

Demis Hassabis, CEO of Google DeepMind, issued a stark warning at the company's developer conference, stating that humanity is standing in the 'foothills of the singularity' and has only a few years to prepare for artificial general intelligence (AGI). Speaking to Axios, he predicted AGI could arrive in as little as four years, reflecting growing industry confidence in the technical path forward. Hassabis emphasized that the upcoming wave of AI agents should be treated as a societal stress test for far more powerful systems. He pointed to Anthropic's Mythos model as a warning shot, illustrating how quickly businesses and governments can be caught off guard by accelerating AI capabilities. He deliberately used provocative language to spark urgency among policymakers and economists, welcoming tentative US government steps toward mandatory safety testing but stressing that safety efforts must accelerate. He also noted ongoing discussions with other top AI labs about safeguards, declining specifics.

Another milestone Hassabis highlighted is recursive self-improvement—systems capable of accelerating their own development. While AI systems do not yet improve themselves independently, he noted 'soft self-improvement' through coding agents that make engineers much more productive. In related remarks, Hassabis criticized companies cutting engineers to ride the AI wave, arguing that when developers become 3-4 times more productive, the smart move is to do more work, not reduce headcount. He offered to hire engineers laid off from other firms, citing a million ideas from lab drug discovery to game design. This message comes during a brutal stretch for tech workers, with 2026 layoffs surpassing 142,000. Hassabis's combined warnings and hiring offer highlight the dual challenge of preparing for AGI while managing current workforce disruptions.