Power, Windows, and Dependencies

Exercise: Map it out

Contested Questions in AI Governance

How should we handle open weights as models get more capable?

When, if ever, should frontier development be slowed - and by whom?

Finding Your Leverage Point

More Career Resources

Next Steps: Fellowships

Next steps: Programs

State of Play at the Frontier

Evaluating Claims

The Evals Theory of Change

METR

Epoch and Jagged Capabilities

The System Card

Independent Investigation: The Hugging Face Incident

Exercise: Briefings

Grappling with the Intelligence Explosion 

Reframing the debate

Exercise: Governance Under Acceleration

Visions of Victory

Face off: Plan S vs Plan A

Argue it out (with yourself)

Start with Radical Optionality as one strategy for acting under uncertainty. Use the IAPS strategic-visions report to place it among alternative pathways, and the CSET reading to uncover the assumptions on which those pathways depend.

Which choices preserve useful options, and which commitments cannot wait? Identify what institutions would need to build before a crisis, who would hold the resulting power, and what evidence would make you change strategy.

The Plan S–Plan A comparison in the next chunk is a focused test of two approaches. It is not the whole governance landscape.


**Plan S vs Plan A**

Assuming international coordination is possible, should governments choose a frontier-development moratorium or verified, safety-conditioned progress?

Complete the shared preparation below: about 25 minutes for Plan A’s summary, FAQ, and verification opening; 20 minutes for the Plan S branch; and five minutes for the independent MIRI treaty summary. If your facilitator assigned you a teaching specialization, also prepare a seven-minute explanation of that plan’s mechanism, hardest dependency, and strongest objection.

**Compare the plans:**
- Trace the route from intervention to a good outcome.
- Identify the hardest dependency and the most consequential failure mode.
- Compare compliance, institutional competence, concentration of power, and conditions for restarting or accelerating development.
- Which institutions or capabilities would both plans need to build before a crisis?
- What evidence would change your own preference?

Both AI 2040 branches come from authors who favour Plan A; the MIRI companion supplies a separate shutdown advocate’s proposal. It is distinct from Plan S. Both plans permit uses of existing AI, and Plan A also includes pauses. 

You may conclude **neither**: identify a premise shared by both plans that you reject and explain how that changes your recommendation. Use the wider IAPS/CSET landscape to develop alternatives. The two steelman exercises below remain focused on your own preferred and least-trusted proposals.


Pick the proposal you find most compelling. Now make the strongest case it creates a dangerous power structure. What could go wrong? Who could abuse the infrastructure it creates?


Steelman 1

Pick the proposal you're most skeptical of. Make the strongest case it's necessary. Under what conditions would you support it despite your reservations?


Steelman 2

Frontier AI Governance

The Proposal Landscape

Race Conditions, Power Concentration and Governance in the Limit

Frontier AI Governance

Visions of Victory

Resources (2 hrs 50 mins)