Why AI Labs Can't Legally Agree to Slow Down

In this episode, James is joined by Sam Hammond, Director of Super Intelligence and Chief Economist at the Foundation for American Innovation and a board member of the AI Alignment Foundation, to discuss what productive AI policy looks like right now, and why the window to act before AI becomes an issue of mass politics is closing fast.

Sam traces how the administration's thinking on AI has shifted since Mythos, from treating AI as a normal technology to taking alignment-style risks seriously. He argues the government needs far better situational awareness of what the labs are building, and that the NSA's 30-day review, focused on cyber and run only at the last stage before release, may have left it knowing less.

The conversation centers on a concrete proposal: Defense Production Act Section 708 agreements that would let frontier labs coordinate on pacing, shared evaluations and alignment research without antitrust liability. Sam explains why China is further behind than the benchmarks suggest, why the real race is between a few US labs, and why slowing down a little could free up the money for an Apollo-scale alignment program.

In this episode:

* Why AI labs can't legally slow down together

* How Defense Production Act 708 agreements could help

* What Washington doesn't know about frontier AI

* Why China is further behind than benchmarks show

* How pacing could pay for alignment research

* What an independent AI safety body could do

* Why AI writing all sounds the same

* The AI Alignment Foundation's neglected approaches

Learn more: https://ae.studio/alignment

AE Studio is hiring: https://www.ae.studio/join-us

Subscribe to our newsletter: https://aestudio.beehiiv.com/

James Bowler LinkedIn: https://www.linkedin.com/in/james-bowler-84b02a100/

Sam Hammond LinkedIn: https://www.linkedin.com/in/samuel-hammond-4b986955