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?

What even are the prospects for international AI governance?

How much direct control should different governments have over AGI?

How should we handle the geopolitical implications of catastrophic dual-use capabilities? 

Tracking Hub

Finding Your Leverage Point

More Career Resources

Next Steps: Fellowships

Certificate

State of Play at the Frontier

Evaluating Claims

The Evals Theory of Change

METR

Epoch

The System Card

Beyond Corporate Promises

Exercise: Briefings

Grappling with the Intelligence Explosion 

Reframing the debate

Exercise: Governance Under Acceleration

Visions of Victory

Face off

Argue it out (with yourself)

Optional but highly recommended

The intelligence explosion has begun. AI capabilities are accelerating faster than any institution can track. Competitive dynamics between states and labs are intensifying. Power is concentrating. 

Pick the governance approach/framing you've found most compelling across Units 1-3 - whether that's a specific proposal like compute governance or RSPs, a broader strategic vision from the IAPS report, or a position you've been developing across the course.

Now walk through specifically what breaks in this world. Be concrete - don't just say "oh, it wouldn't work" identify the first point of failure and the mechanism by which it fails.

Then try to fix it. What would you change about your proposal to make it survive these dynamics? Be specific about the design choices: adaptive mechanisms, sunset clauses, distribution of power, triggers for escalation. Not that you're not writing legislation (AI's will do that soon enough) - here you're trying to identify what features a framework needs to have to function under this kind of acceleration and competitive pressure.

Look at your redesigned proposal through Ngo's lens. Does your fix centralize power in ways that create new risks? If you built stronger oversight, who oversees the overseers? If you distributed power more broadly, can it still coordinate against catastrophic risks? 

Here's a list of things you've encountered so far: 

From Unit 1:
- Safety evaluations / evals-based governance (METR, Epoch, the system card)
- RSPs / ASLs (Anthropic's framework, from reading the system card)
- Third-party evaluation bodies (AISIs, the Board)

From Unit 2:
- International coordination frameworks (International AI Safety Report)
- Common elements of frontier safety policies (METR comparison)
- Policy windows and political dynamics (Leicht)
- China's approach to AI risk (Sheehan/Singer)
- Government preparedness / follow-through (Beers)

From Unit 3:
- Compute governance (Heim et al.)
- Private governance / tort liability (Ball)
- Nine strategic visions across two axes (Delaney et al.)
- Moderate institutional regulation (Anderljung, in go-deeper)
- Foundational assumptions underlying proposals (CSET method)

From Unit 3 strongly recommended:
- The narrow corridor / constitutional constraints on government AGI control (Bullock/Hammond/Krier)
- Offense-defense balance / defensive AI deployment (Krier)


Finally, produce a one-page brief titled "What [your chosen proposal] needs to become to survive the intelligence explosion - and what that costs." 


Frontier AI Governance

Race Conditions, Power Concentration and Governance in the Limit

The Proposal Landscape

Frontier AI Governance

Exercise: Governance Under Acceleration

Resources (15 mins)

Exercises