Scaling Laws Rapid Response: An "FDA for AI" at the White House?, with Dean Ball
May 8, 2026
Dean Ball, former White House AI policy advisor and current senior fellow focused on frontier AI governance. They trace how Anthropic's Mythos jolted Washington, debate reported plans for pre-release model vetting, weigh voluntary "kick the tires" frameworks versus mandatory licensing, and explore how oversight could scale as model capabilities accelerate.
AI Snips
Chapters
Transcript
Episode notes
Mythos Sparked A White House Policy Pivot
- Mythos and similar frontier models triggered a rapid policy pivot because their cyber capabilities made risks suddenly concrete for senior officials.
- Dean Ball says the Anthropic demonstration brought new, powerful actors into the fold who previously viewed AI more optimistically.
White House Lacks Clear Authority For Mandatory Vetting
- There is no clear legal authority for the White House to impose a mandatory pre-deployment vetting regime absent new legislation.
- Alan Rozenshtein points to the Defense Production Act as limited precedent but emphasizes Youngstown-style limits on unilateral executive control over private firms.
Use Voluntary "Kick The Tires" Testing First
- Use voluntary testing channels like CAISI plus the Cyber Resilience Fund to build a "kick the tires" pre-deployment period that helps local partners shore up defenses.
- Kevin Frazier recommends expanding CAISI labs' testing and using DHS funding to strengthen state, local, and tribal readiness.