Overview
The DeepMind Institute (DMI) is a public essay platform launched by google and google-deepmind on September 17, 2026 to broaden debate on artificial-general-intelligence safety, policy, philosophy, and societal impact. Essays reflect authors’ views and are explicitly not Google’s official position.
Leadership
- Managing Editor: shane-legg (Chief AGI Scientist, Google DeepMind co-founder)
- Directors: demis-hassabis (DeepMind chair, Alphabet chief scientist), james-manyika (Google president of research, labs, technology and society)
Inaugural Publications (September 2026)
| Essay | Authors | Topic |
|---|---|---|
| The case for reasoning transparency | Rohin Shah, Anca Dragan | Preserving human-readable model reasoning; limits on “opaque serial depth” |
| Economic policy for AGI | Julian Jacobs, Alex Imas | Managing potential AGI economic disruption |
| Principles for a new utopianism | Stephen Cave | Human flourishing principles |
| A framework for frontier AI | Demis Hassabis | Frontier model evaluation; U.S.-led standards body proposal |
Policy Positions
- Hassabis proposal: Voluntary then mandatory pre-deployment reviews for frontier models; potential coordinated slowdowns among frontier labs (frontier-ai-governance)
- Legg on pacing: Called dario-amodei’s “pace not pause” frontier release proposal “interesting directionally” and “worth considering” (pace-the-frontier-framework)
- AGI timeline: Legg remains comfortable with 50% chance of “minimal” AGI by 2028; called recent AGI arrival claims premature
Hassabis's standards body and mandatory review proposals are policy suggestions, not enacted regulation. Note potential conflict of interest — frontier lab proposing self-regulation framework.