Recovery Blueprint: AI Governance in Bilateral Negotiations

The Structural Problem

When President Trump announced he had discussed artificial intelligence guardrails with Chinese President Xi Jinping, the immediate reaction focused on the politics of the conversation. But the underlying structural problem is more fundamental: the United States currently has no clear institutional mechanism that defines who has authority to negotiate international AI standards, what processes constrain those negotiations, or how such commitments become binding on domestic actors.

This is not a problem unique to Trump or to China policy. It is a design gap in American governance that exists because AI regulation has emerged faster than the constitutional and statutory frameworks designed to manage technology policy. The executive branch possesses broad foreign affairs powers. Congress holds authority over interstate commerce and can regulate technology. Independent agencies like the Federal Trade Commission have sector-specific mandates. But when a president discusses "guardrails" with a foreign leader, which institution actually has the authority to implement those guardrails? What prevents a president from making informal commitments that bypass congressional review, agency expertise, or public input?

The current arrangement allows executive discretion without structural accountability. That creates two cascade risks: first, that informal bilateral agreements become de facto policy without democratic legitimacy; second, that foreign powers exploit this ambiguity by negotiating commitments with executives that cannot or will not be honored by successor administrations or coordinate agencies.

Root Cause: Jurisdiction Without Architecture

The root cause is not executive overreach in the traditional sense. It is jurisdictional ambiguity combined with procedural absence. The Constitution divides foreign policy and domestic regulation into separate domains, but AI governance sits at the intersection. Is an agreement on AI safety standards a treaty requiring Senate ratification? An executive agreement under existing statutory authority? A non-binding political commitment? The answer depends on how the conversation is characterized, and there is no institutional mechanism that forces clarity before the conversation happens.

Moreover, no statute currently defines what qualifies as a binding AI policy commitment or what interagency process must review such commitments before they are made. The result is that a president can discuss "guardrails" without triggering any mandatory consultation with the Department of Commerce, the National Institute of Standards and Technology, the Federal Trade Commission, or Congress — even though each of these entities has independent statutory authority over aspects of AI development and deployment.

This is a structural failure of process design. The system assumes that foreign policy and domestic regulation will be conducted in separate channels with separate authorities. It has no load-bearing mechanism for managing technology negotiations that inherently merge both domains.

Calibration 1: Statutory Definition of AI Policy Commitments

What it changes: Congress must enact legislation that defines what constitutes a "binding AI policy commitment" and establishes a mandatory pre-negotiation review process. The statute should specify that any bilateral or multilateral discussion involving technical standards, safety protocols, data governance, or export controls related to artificial intelligence systems triggers a requirement for the president to submit a preliminary negotiation framework to the House Committee on Energy and Commerce and the Senate Committee on Commerce, Science, and Transportation at least 30 days before substantive negotiations begin.

Who implements it: Congress, through ordinary legislation. This does not require a constitutional amendment and does not infringe on the president's foreign affairs powers — it simply defines the conditions under which certain commitments become binding on domestic agencies and actors.

What it repairs: This Calibration creates procedural friction that forces clarity. If a president wants to make a commitment that will bind American AI developers or regulators, that commitment must be defined in advance and reviewed by the legislative branch. If the president wants to have a non-binding political conversation, that is permissible — but it cannot later be converted into binding policy without returning through the defined process. The repair is not a veto over foreign policy; it is a structural requirement for transparency and coordination.

Calibration 2: Interagency AI Negotiation Coordinator

What it changes: Establish by executive order a permanent Interagency AI Negotiation Coordinator, housed within the National Security Council but with a statutory reporting requirement to Congress. This coordinator must be included in any bilateral or multilateral discussions involving AI policy and must prepare a written assessment of how any proposed commitment aligns with existing statutory mandates of the FTC, Department of Commerce, and other agencies with regulatory authority over AI.

Who implements it: The president, via executive order, with Congress codifying the reporting requirement and funding through appropriations. This is a hybrid mechanism — the executive creates the coordinator, but Congress ensures it has structural independence through mandatory reporting.

What it repairs: This creates a single institutional actor responsible for managing jurisdictional overlap. Currently, no one is tasked with ensuring that presidential commitments in foreign negotiations are compatible with domestic regulatory architecture. The coordinator does not have veto power, but their written assessment becomes part of the legislative and public record. If a president makes a commitment that contradicts the coordinator's assessment, Congress and the courts have a clear basis for evaluating whether the commitment exceeds executive authority.

Calibration 3: Conditional Enforceability Provision

What it changes: Amend the Administrative Procedure Act to specify that any federal agency action implementing an international AI commitment made without following the procedures in Calibration 1 is subject to expedited judicial review under an "ultra vires" standard. Plaintiffs challenging such agency action would not need to demonstrate standing under traditional injury-in-fact tests; instead, any entity subject to the policy could petition for review on the grounds that the policy lacks proper statutory or constitutional foundation.

Who implements it: Congress, through amendment to 5 U.S.C. § 706 or related provisions of the APA. Federal courts would then apply this standard in reviewing challenges.

What it repairs: This Calibration prevents the back-door enforcement problem. Even if a president makes a commitment without following proper procedure, that commitment cannot be operationalized through agency rulemaking without triggering judicial scrutiny. This does not prevent the president from conducting diplomacy, but it prevents informal diplomacy from being converted into binding domestic regulation without structural accountability.

Feasibility and Minimum Viable Repair

Of the three Calibrations, the second — the Interagency Coordinator — is the most immediately achievable. It can be established by executive order and does not require overcoming legislative gridlock. It provides immediate structural improvement by ensuring that at least one institutional actor is tasked with coherence.

The minimum viable repair to prevent cascade failure is Calibration 1. Without it, the jurisdictional ambiguity will continue to allow informal commitments to migrate into binding policy without democratic input. The risk is not that any single conversation will produce catastrophic policy; the risk is that the absence of process will erode the legitimacy and stability of AI governance itself, making it impossible for foreign partners, domestic developers, or the public to know what commitments are real and enforceable.

The system does not need to prevent presidential diplomacy. It needs to define when diplomacy becomes policy — and ensure that transition is visible, accountable, and structurally sound.