A major artificial-intelligence vendor declined to allow its systems to power autonomous lethal weapons or mass surveillance targeting Americans. Rather than simply switching suppliers, federal authorities classified the company as a "supply-chain risk to national security"—terminology ordinarily reserved for hostile foreign powers—and systematically excluded it from the entire defense contracting apparatus. A presidential order then instructed all federal agencies to abandon the company's products.
This sequence represents far more than a routine business disagreement. It exemplifies what legal scholars term "expressive governance": the deployment of standard administrative mechanisms—contracts, grants, licenses, hiring decisions, regulatory classifications—to suppress disfavored opinions while maintaining a veneer of procedural legitimacy. A forthcoming book and an amicus brief filed in Anthropic v. US Department of War, currently litigated in federal court in San Francisco, examine this phenomenon in detail.
The peril lies in expressive governance's superficial legality. Judicial precedent has long, and reasonably, granted executives broad discretion over contract awards and security clearances. Yet this deference assumed genuine operational decisions, not retaliatory measures. When authorities deploy these inconspicuous instruments against speakers, constitutional breaches can remain concealed within the space separating official justifications from actual motivations.
The AI company is hardly an isolated case. Identical tactics have targeted legal practices denied federal access for defending unpopular clients, educational institutions threatened with funding withdrawal over campus speech policies and faculty viewpoints, news outlets facing regulatory sanctions, and government employees terminated for co-signing internal dissent memoranda. Federal courts have already struck down numerous such actions as unconstitutional. In every instance, the mechanism proved consistent: formally neutral authority weaponised to inflict tangible financial and reputational damage against those refusing compliance.
The cumulative damage transcends individual cases. Every corporation, partnership, and institution relying on federal contracts, grants, or regulatory approval now internalises a stark warning: articulate a principled public position the sitting administration opposes, and you risk becoming the next target. As Luke Barnes noted in analysis for the NYU Stern Center for Business and Human Rights, the message reverberates across entire sectors: ethical responsibility transforms into commercial liability. This silence—positions never articulated, stands never taken—represents the structural suppression the First Amendment was designed to prevent.
A Framework for Courts
The critical challenge involves determining how courts should respond. Invalidating individual orders produces minimal protection when the same tactic simply resurfaces under different administrative guises. The amicus brief in the Anthropic matter proposes a structured approach furnishing judges with consistent methodology rather than isolated remedies. The framework rests on three components.
Clear Congressional Authorization
Authorities must identify specific statutory language permitting their conduct—a requirement for explicit legislative direction. Before executives exercise discretionary authority in ways burdening expression, courts should mandate that Congress explicitly sanctioned that particular application. A classification intended for foreign sabotage cannot be repurposed to penalize an American company's policy positions without express congressional instruction. No such authorization exists in this instance.
Reversed Burden of Justification
Once a complainant demonstrates both a pattern connected to viewpoint—questionable timing, selective enforcement, incriminating public remarks—and an exploitable dependency such as a contract or license, the government must substantiate that it would have acted identically for legitimate reasons. Should the genuine motivation involve the speech itself, officials cannot obscure this behind bureaucratic language.
Protection for Expressive Intermediaries
The presumption gains particular weight when authorities target what scholars call an "expressive intermediary." This designation differs from Section 230 platform liability; it describes entities whose own communicative decisions fundamentally influence how broad populations locate, evaluate, and disseminate information, such that targeting them distorts public discourse beyond the immediate victim. Large language models exemplify this category: an AI company's publicly articulated safety commitments constitute its own protected expression, and its systems increasingly function as intermediaries through which millions of people and institutions consume information. When government acts against such entities, courts should presume the action warps public debate; vague invocations of "national security" prove insufficient, and authorities must instead furnish concrete, content-neutral, case-specific justifications.
This approach introduces no novel constitutional standard and does not constrain executive authority. Government retains complete freedom to dispute a company's positions, decline its proposals, and redirect procurement elsewhere. What remains prohibited is conscripting national-security apparatus to demolish a business for expressing views on matters of genuine public concern. Courts equipped with proper analytical tools can distinguish between these scenarios.
The implications extend far beyond a single corporation or administration. Should expressive governance persist unchecked, principled conviction becomes economically catastrophic, and space for candid public discourse gradually evaporates. Courts occupy the institutional position to reject this trajectory—and must do so.
Source: Tech Policy Press
