Public Policy & Governance

House negotiators are discussing a provision that would preempt specific state AI laws, with California and New York explicitly named as targets, according to Politico. This is not aspirational rhetoric — it represents active legislative drafting in the context of broader AI legislation, and its inclusion would mark the most consequential federal intervention into AI governance since the current Congress convened. The discussions also include model vetting requirements, suggesting the package combines preemption with affirmative federal standards.

The strategic logic from the industry's perspective is transparent: California's AI legislation has historically set de facto national standards, and New York has moved aggressively on algorithmic accountability. A federal preemption floor — even one with relatively light-touch requirements — would neutralise both. The complication is that preemption provisions have historically been the most contentious element of federal tech legislation, with state attorneys general and civil liberties coalitions reliably mounting constitutional challenges. The Mythos situation appears to be providing political urgency that might otherwise be absent, with the White House now invested in demonstrating federal-level control over frontier AI.

The Institute for AI Policy and Strategy has published a detailed policy playbook titled 'After Mythos' explicitly addressed to White House decision-makers, proposing a suite of executive actions calibrated to frontier AI national security risks IAPS. The significance here is the framing: this is not a paper proposing future regulatory structures but a document responding to a specific, apparently ongoing situation that has already triggered White House deliberations. IAPS is a credible policy shop with direct government engagement, and publications of this type typically signal active back-channel advisory relationships.

The national security framing of frontier AI governance has institutional consequences that purely commercial or consumer-protection framings do not. When AI risk is classified as a national security matter, the relevant actors shift — from FTC and NIST to NSC, ONCD, and potentially DoD — and the procedural constraints on executive action narrow considerably. Executive orders in the national security space face different judicial review standards than commercial regulation. This is the most significant structural shift in AI governance authority currently in motion, and it is being driven by executive branch actors rather than Congress.

The Center for Democracy and Technology's second installment in its AI in Policing series examines gunshot detection technology — acoustic AI systems deployed in dozens of US cities — and documents the gap between vendor claims and evidentiary performance, as well as the near-total absence of formal accountability mechanisms governing their use CDT. This is a domain where the implementation gap is at its starkest: the technology is operational and consequential, generating police dispatches that have in documented cases led to wrongful stops and arrests, while federal regulatory frameworks do not address it and most municipal deployments lack formal use-policy requirements.

CDT's positioning here matters because they are a credible interlocutor with both Congressional staff and executive agency officials. Publications in this series tend to feed directly into Congressional testimony and agency guidance development. The absence of any federal framework — FTC jurisdiction is contested, NIST's AI RMF is voluntary, and there is no law enforcement AI-specific statute — means this is an area where state legislation or municipal ordinance represents the only viable near-term accountability mechanism. Several cities including Chicago and San Francisco have enacted or proposed algorithmic accountability ordinances, but enforcement track records are thin.