Washington group borrows nuclear and banking rules to govern AI

Silicon Valley loves to claim that lawmakers in Washington have no idea how software works and shouldn't touch AI.

Now, a new group inside the Beltway is pushing back with a simple pitch: you don't need to reinvent the wheel to regulate frontier AI, you just need to borrow the playbooks already used for banks, cybersecurity, and nuclear reactors.

Borrowing blueprints from Wall Street and the Cold War

Americans for Responsible Innovation (ARI), a new bipartisan advocacy group, launched on Wednesday to bridge the gap between AI development and federal policy.

The group is led by former Congressman and Army Under Secretary Brad Carson alongside tech entrepreneur Eric Gastfriend. Its advisory bench features Stanford Digital Economy Lab director Erik Brynjolfsson and Diane Rinaldo, former acting chief of the National Telecommunications and Information Administration.

Carson saw AI's military implications while at the Pentagon, but said ChatGPT made him realize civilian agencies were dangerously unprepared.

Instead of waiting years for Congress to pass an entirely new digital regulatory regime, ARI wants to build on President Biden's AI executive order by copying proven shortcuts from high-risk sectors.

Their proposal starts with creating an AI Auditing Oversight Board, modeled directly after the accounting oversight body Congress built for public companies after Enron.

To control compute, ARI wants banking-style Know Your Customer rules that restrict who can rent American cloud infrastructure to train advanced frontier models.

On the hardware and supply chain side, they propose a democratic AI Suppliers Group modeled after the Cold War-era Nuclear Suppliers Group.

They are also calling for mandatory incident reporting databases borrowed from cybersecurity, alongside a massive funding boost for the Commerce Department's National Institute of Standards and Technology (NIST) to anchor federal oversight.

Lobbyists, budgets, and the software precedent

Carson argues that while pharmaceuticals, commercial jets, and nuclear reactors are all unique, the regulatory tools governing them are adaptable.

He also pointed out that Big Tech cannot be trusted to represent the public interest due to obvious commercial conflicts.

Still, ARI faces brutal odds on Capitol Hill.

Congress has literally never passed comprehensive federal regulations for software or digital platforms. Silicon Valley's lobbying apparatus is massive, and many AI startups are stacked with tech veterans who spent the last decade dodging platform rules.

Money is an immediate hurdle too. While ARI wants the Commerce Department to lead the charge, the agency is currently facing federal budget cuts this year.

Why it matters

If ARI's proposals gain traction on the Hill, the downstream friction for AI builders will be immediate.

Applying banking-style Know Your Customer rules to GPU clusters means cloud providers will have to aggressively vet anyone spinning up large-scale training runs. That would dramatically raise compliance costs and bureaucratic hurdles for startups and independent researchers trying to access compute.

From an industry perspective, we think ARI's strategy is undeniably clever. They are dodging the impossible debate over passing a grand AI manifesto through Congress, focusing instead on pragmatic, sector-tested mechanisms that existing federal agencies can execute tomorrow.

However, the core risk is false equivalence. Nuclear material and bank ledgers are physical and centralized; weights, algorithms, and open-source models move across borders in seconds. Treating code like enriched uranium might sound reassuring in a Senate hearing, but it risks burdening domestic builders while doing very little to slow down competitors overseas.