Brookings Urges Federal AI Law to Replace 'Black Box' Executive Rule
A new proposal calls for mandatory audits and licensing for frontier models to address material risks and end regulatory volatility.
The Brookings Institution is calling on Congress to enact a comprehensive federal law for AI governance to replace the current system of volatile, ad hoc executive-led decisionmaking. The proposal argues that relying on executive orders and voluntary frameworks creates instability for the industry and fails to provide a durable regulatory foundation.
At the center of the proposal is a requirement for mandatory third-party audits and licensing for frontier AI models. This approach mirrors a growing trend at the state level; Brookings notes that Illinois has already enacted the first mandatory annual requirement, while California and New York are adopting similar standards. The proposal specifically advocates for a regulatory focus limited to "material" risks, which include automated cyberwarfare and infrastructure disruption, the design of novel viral pathogens or bioweapons, and the potential for harmful autonomous action resulting from a loss of control.
Closing Regulatory Loopholes
Brookings argues that current legislative drafts, such as the Great American AI Act, contain a critical flaw: the distinction between "developers" and "deployers." According to the Brookings author, this distinction is a "structural loophole that invites regulatory evasion." The proposal asserts that accountability should instead fall on whichever entity provides the model to the public, regardless of whether they built it or simply deployed it.
Furthermore, the proposal suggests a targeted approach to federal preemption. It recommends that the federal government take sole authority over existential risks—specifically cybersecurity and biological threats—while leaving state-level privacy and consumer protection laws intact to maintain local safeguards.
National Security and Stability
This push for statutory governance comes amid shifting international diplomacy, highlighted by the June 2026 G7 summit in Évian, France, where AI governance was a central theme. A primary concern for Brookings is the current lack of transparency regarding export controls. The author argues that export control authority should not be a "black box" that operates in parallel to AI governance, urging Congress to reconcile these powers to prevent arbitrary actions by the executive branch.
The Path Forward
Establishing a statutory framework would move the U.S. away from secretive executive licensing toward a public, accountable system that incorporates civil society input. By doing so, the U.S. could balance national security objectives with innovation while avoiding a fragmented landscape of conflicting state laws. The success of this transition depends on whether Congress is willing to strip the executive branch of its current discretionary power and codify a risk-based standard for the most powerful AI systems.