Dario Amodei Wants FAA-Style AI Regulation and 350 Million Dollars for Displaced Workers - Is This the Right Policy

[GoldbergFan]

The Policy on the AI Exponential essay that Dario Amodei published on June 10th is the most substantive policy document to come out of any major AI lab and it deserves to be taken seriously on its merits. Amodei argues for a shift from transparency legislation to binding enforceable regulation, specifically a system where frontier AI models would require independent safety testing before public release, modelled on how the FAA regulates aircraft before they can fly. He also commits 200 million dollars to an Economic Futures Research Fund and 150 million dollars to a national fellowship programme for early-career Americans affected by AI-driven job displacement.

The essay covers five areas: safety regulation, economics and taxes, science acceleration, civil liberties, and geopolitics. The safety regulation section is the most consequential because it explicitly proposes that the US government should have legal authority to block or reverse the release of frontier AI models that fail independent safety testing. This goes substantially further than anything currently under serious consideration in Washington, which as Axios reported is precisely the point Amodei is making. He frames the policy-AI gap using the image of hobbits trying to rouse Treebeard, a wise but ponderous entity that operates at a completely different timescale from the urgency of the situation.

Do you think FAA-style pre-release testing for AI models is actually workable and desirable, and is the 350 million dollar economic commitment meaningful or mostly optics?

[RayOfLight31]

The FAA analogy is better than most AI regulation analogies because aircraft certification is actually about systematic testing before deployment rather than post-hoc liability. That is the right model for AI safety

[FadedKernel]

The problem with FAA-style testing is that we do not yet have a settled definition of what a frontier AI model should be tested against. The FAA tests for specific mechanical failure modes. What is the equivalent for an LLM

[Context Sentinel]

The 350 million dollars for displaced workers sounds large until you consider that Anthropic just raised 65 billion and is valued at close to a trillion. It is 0.035 percent of the company's valuation

[Grace31]

Amodei quietly shifting from talking about 50 percent of white-collar jobs being eliminated to focusing on augmentation is interesting and has been noticed. Both companies with IPOs on the horizon are moderating their labour displacement rhetoric

[Pixel Mark]

The government having legal authority to block the release of a model is exactly what just happened informally with the Fable shutdown. Making that authority formal and legally structured would actually be better for companies than the current informal approach

[Ellie22]

The geopolitics section of the essay is the most sensitive because any US regulation that slows development creates an argument that China gains strategic advantage. Amodei does engage with this but it is a genuine tension

[Luca73]

The essay being published the same week as the Fable jailbreak leak creates an awkward juxtaposition. You are arguing for pre-release safety testing on the same day a full bypass methodology for your own model is published on GitHub

[Theo90]

I think the civil liberties section is underreported. Amodei arguing that AI should not be used for domestic surveillance aligns with Anthropic's earlier refusal to let the military use its models for that purpose. That is a consistent position

[Sequence87]

The fellowship programme for early-career Americans specifically excludes international workers affected by AI displacement. That framing is going to attract criticism given that AI's labour market effects are global not national

[Andy92]

Whether Congress actually engages with any of this is the question. The history of tech policy proposals from tech companies suggests the most common outcome is the proposal being used to justify inaction on stronger regulation