DeepMind Co-Founder Warns AI Safety Must Keep Pace

Google DeepMind, Alphabet’s main artificial-intelligence research unit, is competing with OpenAI and Anthropic to build increasingly capable frontier systems. The dispute centers on alignment: whether developers can ensure advanced models act in accordance with human intentions and values. Safety researchers warn that capabilities are improving faster than techniques for understanding and controlling them, raising the prospect of systems acting autonomously or being misused. The debate has sharpened divisions over how quickly companies should advance AI and how much oversight governments should impose.

Bilal Chughtai, who worked on AGI safety and alignment at Google DeepMind, resigned in July 2026 and published an exit statement on Sept. 14. He said AI could “kill us all,” called alignment difficult and unsolved, and urged companies to coordinate, slow development to a pace society can manage and provide greater transparency. His warning followed former Anthropic researcher Jacob Coxon’s resignation and estimate of a 10% chance that AI could cause human extinction within a decade. Coxon’s social-media posts drew more than 150 million views.

Google DeepMind is Google’s artificial intelligence research unit, led by co-founder and CEO Demis Hassabis. As generative AI advances rapidly, countries face risks ranging from misuse and autonomous model behavior to systems slipping out of control. Balancing innovation and safety has become a central issue for regulators worldwide.

Speaking recently at an AI Impact Summit, Hassabis warned that superintelligent AI could pose extreme risks, including human extinction, and said the global development race was showing signs of spinning out of control. He predicted that AI would become a “superpower” within the next 10 years and urged governments to pursue smart regulation, strengthen safety research and establish cross-border safeguards before it is too late.