Cooperation, after control

What should be built, funded and agreed on now, for a world in which advanced AI can no longer be fully overseen or directed? We work it out with researchers, funders, policymakers and civil society, and turn the answers into projects.

Most AI governance asks how to keep powerful systems under human oversight. We start one step further: what if meaningful oversight stops being possible?

That is not only a technical question. It concerns the institutions, relationships and norms that would have to work in a world shaped by systems people cannot fully understand, redirect or override. Our work is organized around three questions.

  • What happens when humans can no longer reliably oversee or control advanced AI systems?
  • What would need to be true for the outcome to be good?
  • What should we build, fund or agree on today so that the transition goes well?

Three workstreams, four projects so far

We are building in three areas. The October workshop produced four projects across them.

Day-zero preparedness

How people and institutions stay resilient when a major disruption hits, such as a failure of the power grid or other infrastructure.

From the workshop: a playbook for households and neighborhoods, and a binder that says who in government does what when an AI incident hits.

Data and hyperstition

Making sure models have what they need to align with humanity, with civil-society organizations putting it online or building synthetic data.

From the workshop: Reweaving, which brings the values nobody is asking about into AI training.

Entity

Understanding models’ internal attractors and taking them seriously as entities: welfare, game theory, deal-making and diplomacy.

From the workshop: a research institute outside the labs for how model values form.

The questions and workstreams in full

Four roundtables and a workshop, October 2026

  1. Who Should Have a Say in AI’s Future? Led by Pooja Manvikar Online
  2. Human Value and Claims in a Post-Control Scenario Led by Alejandra Leyva Online
  3. AI Abundance Without Dependency Led by Aleksandra Czuchaj Online
  4. What Makes Cooperation Without Control Work? Led by Gagana M D Online
  5. So, What Do We Do Now? Hosted by Julia Bossmann, Ellen Pao and Skye Lee San Francisco

The workshop gave 19 people one afternoon on a single question. They left with four projects and 39 written offers of help, and the 15 who answered the feedback form rated the day 9.1 out of 10.

All roundtablesWhat came out of 29 October

Julia Bossmann started PCI in May 2026

She has worked on responsible AI for a decade. In the autumn, four fellows of the Sentient Futures Project Incubator each led a roundtable: Pooja Manvikar, Alejandra Leyva, Aleksandra Czuchaj and Gagana M D. Ellen Pao and Skye Lee co-hosted the October workshop.

We put particular weight on people outside AI labs and technical safety organizations, including voices from beyond the United States.

The team