Google launches DeepMind Institute to explore AGI deployment

Google DeepMind has launched the DeepMind Institute, led by Shane Legg, Demis Hassabis, and James Manyika, to explore the societal and policy impacts of artificial general intelligence (AGI). The launch coincides with warnings from Legg that AI capabilities must not outpace safety controls. While some competitors claim AGI has arrived with models like OpenAI's GPT-6 Astra, DeepMind executives dispute this, maintaining that true human-level cognitive AGI remains years away.

DeepMind Institute launch

• ▪The DeepMind Institute is led by managing editor Shane Legg alongside co-directors Demis Hassabis and James Manyika
• ▪Google DeepMind launched the DeepMind Institute on September 16, 2026, to explore the societal, scientific, and policy implications of artificial general intelligence
• ▪The DeepMind Institute launched with inaugural publications covering topics such as economic policy for artificial general intelligence, model reasoning transparency, global access, and human flourishing

AGI timeline predictions

• ▪Google DeepMind Chair Demis Hassabis projects that artificial general intelligence could be achieved within a few years, while OpenAI CEO Sam Altman believes it could arrive by the end of 2026
• ▪Google DeepMind co-founder Shane Legg stated on September 16, 2026, that he remains comfortable with his long-held forecast of a 50 percent chance of achieving minimal artificial general intelligence by 2028

AGI definition disagreements

• ▪Google DeepMind defines artificial general intelligence as a system that exhibits all the cognitive capabilities of the human brain
• ▪Shane Legg rejected claims by OpenAI President Greg Brockman and Nvidia CEO Jensen Huang that OpenAI's GPT-6 Astra model demonstrates that the artificial general intelligence era has arrived

AI safety controls

• ▪An essay by DeepMind researchers Rohin Shah and Anca Dragan criticized the speedy release of OpenAI's GPT-6 Astra model, claiming it showed a downward trend in transparency
• ▪Shane Legg warned on September 16, 2026, that rapidly advancing artificial intelligence capabilities must not outrun safety controls

Global AI collaboration

• ▪Google Senior Vice President James Manyika stated that artificial intelligence safety and responsibility must be addressed globally rather than on a country-by-country basis
• ▪Shane Legg stated that scientific breakthroughs in artificial intelligence advance a global community, making international collaboration essential

Debatable claims

• ▪Governments should mandate transparency in AI model reasoning
• ▪AI safety efforts should focus on international diplomacy rather than unilateral national controls
• ▪AI developers should slow down the release of frontier models