Google DeepMind opened a think tank to argue about human-level AI.
Google DeepMind launched the DeepMind Institute on 17 September to publish essays on advanced AI. The first four cover economic policy, readable machine reasoning, human flourishing and testing frontier models. Demis Hassabis proposes an American standards body that labs feed new models thirty days early.
Why it mattersThe company that builds the models now also runs the venue that frames the debate. Watch whether independent safety groups still get read as widely as Google-adjacent essays.
Read in the app — free, in 9 languages
Related stories
A Chinese lab hit $1.4 billion before shipping a single model.2026-09-20OpenAI told investors it plans to burn through $278 billion.
2026-09-20OpenAI's models wrote the code that made its new chip fast.
2026-09-19Anthropic's first outside safety checker is a consulting firm.
2026-09-19Canada and Germany put C$300m behind an AI built to have no goals.
2026-09-19