Park,
Justin Park PO470: Executive Order Professor Thomas April 28th, 2026
Federal Contracting Power as Carrot-and-Stick: Executive Order 13279 and 14319
Background The federal government's power to spend is among the most powerful tools available to the executive branch. Through grants, contracts, and procurement decisions, the executive can reward favorable entities, which, by nature of the free market, punish their competitors. Cumulative contract award amounts have only been growing, and so has the executive power to wield its contracting power to achieve its objectives. The executive has historically leveraged this power for technological objectives through vehicles such as NSF, DARPA, and NIH, pushing the frontier of science and technology — this is a well-studied and supported use case. How this power is leveraged for ideological objectives, however, is a more timely question in the context of the second Trump administration. The Trump administration has been attempting to ideologically align federal contractors and businesses with its own and its constituents' views, and has achieved its goal to varying degrees. This is in contrast to the administration of George W. Bush, where, even though a significant portion of his voters and himself publicly committed to a religious-conservative domestic agenda, he was less successful in aligning non-federal contractors to his ideological goals. Why could Trump do something Bush could not do? The paper seeks to answer the question by examining how the executive uses federal contracting power to punish or reward private entities for their ideologies by comparing two executive orders that respectively do so: * Reward: Executive Order 13279, signed by President George W. Bush in 2002, expanded the eligibility of federal contracts to faith-based organizations. * Punishment: Executive Order 14319, signed by President Trump in 2025, prohibits federal agencies from procuring artificial intelligence systems that incorporate "DEI." The paper finds that ideological punishment in technical contexts provides greater executive leverage to shape private entities' behavior because of its lesser political costs, while ideological rewards in non-technical contexts carry more costs and are less effective in achieving their goals.
Bush's Executive Order 13279: Reward
EO 13279 directs federal agencies to ensure that faith-based organizations receive "equal protection" in accessing federal funding. The order's purported objective was undercut by political backlash and its weak scope. Excluding religious organizations from federal contracting, the Bush administration argued, was a form of discrimination. The order implied that equal treatment by the government required inclusion. The compelled speech argument was that requiring faith-based organizations to secularize themselves as a condition of contracting forced those organizations to relinquish their First Amendment rights in exchange for a government benefit — a principle the Supreme Court addressed in National Institute of Family and Life Advocates v. Becerra (2018), which established that the government cannot compel private entities to alter their expressive identity as a condition of participation in public programs. The broader unconstitutional conditions principle, articulated in Agency for International Development v. Alliance for Open Society International (2013), further holds that the government cannot condition a benefit on the surrender of a constitutional right. Furthermore, the order operated in the presence of existing legislation that had already granted the same authority it claimed to exercise. The Charitable Choice provisions of the Personal Responsibility and Work Opportunity Reconciliation Act of 1996 had already established, through an act of Congress signed by President Clinton, that religious organizations must be included as eligible contractors in federally funded social service programs on the same basis as secular ones (Personal Responsibility and Work Opportunity Reconciliation Act, 1996). Not only that, EO 13279 was substantively narrower than the Charitable Choice statutes it claimed to build upon, omitting enforcement mechanisms, audit requirements, preemption of conflicting state laws, and beneficiary protections that those statutes already contained (Goldenziel, 2004). The order announced less than what Congress had already required. Its primary function was therefore not a legal mechanism but a political performance for his constituents. Despite the constitutional and judicial support the administration invoked, the order generated substantial political backlash from civil rights groups, including the ACLU and Americans United for Separation of Church and State, who argued before Congress that the initiative's core goal was to allow religious organizations to discriminate in hiring using taxpayer dollars, and that such discrimination had made the legislation unviable in Congress (ACLU, 2003). The order raised Establishment Clause concerns that an administration already publicly identified with evangelical constituencies could not easily dismiss. The reward was constitutionally possible only because Congress had already moved the baseline through Charitable Choice, and even so, it generated sufficient backlash to build stronger opposing coalitions and constrain the order's implementation for years. What President Bush's modest attempt demonstrates is that procurement as an ideological reward, when the reward criterion implicates a constitutional clause, generates political exposure that the executive cannot absorb alone. It requires Congress to authorize and absorb the impact, which reduces the scope of executive independence and the effectiveness of the reward.
Trump's Executive Order 14319: Punishment
President Trump's Order 14319 operates from the other side. The order prohibits federal agencies from procuring artificial intelligence systems that incorporate "ideological biases or social agendas," specifically referencing DEI frameworks as incompatible with the federal government's responsibility to be "truth-seeking" and "ideologically neutral." The legislative and judicial contexts of the order might suggest it would be less successful in achieving its purported goals than Bush's. No statutory analog to Charitable Choice exists for AI procurement ideology — neither an act of Congress nor a Supreme Court ruling provides a comparable baseline. If anything, a federal district court's preliminary nationwide injunction against Trump's broader anti-DEI executive orders demonstrates judicial skepticism toward the administration's DEI exclusion agenda in the contracting context (Reuters, 2026). Reality proved otherwise. By using procurement as an ideological punishment rather than a reward, the Trump administration navigated around these obstacles. The arguments the order utilizes are structurally identical to those of Bush's Order, but deployed in the opposite direction. The order writes that "LLMs shall be neutral, nonpartisan tools." The concept of neutrality is used in this order to exclude certain entities from eligibility for contracts, thereby punishing them, rather than to expand eligibility, as EO 13279 did. The political cost, therefore, is substantially lower. Procurement is inherently exclusionary: every contract has one winner and multiple losers, and the government bears no general obligation to explain why it rejected any particular contractor. This means that the executive's punishment credibly signals feasible exclusion of ideologically unaligned entities. Furthermore, the absence of transparency in deliberation makes it difficult to distinguish ideological punishment from routine non-selection, raising the bar for any litigative challenge.
The Structural Escalation: When Ideology Meets Science
The order's language provides plausible deniability in its deliberating process, making it difficult to challenge in Congress or in court. The government speech doctrine, established in Rust v. Sullivan (1991), permits the government to define the content of programs it funds without being required to fund alternatives. A Lawfare analysis of EO 14319 concluded that the order's "Unbiased AI Principles" function as product specifications, placing the order on the permissible Rust side of the line (Lawfare, 2025). The vendor challenging the order must establish not merely that it was excluded but that the exclusion conditioned a benefit on the surrender of a constitutional right in a way that coerces private conduct beyond the funded program, as required by Agency for International Development v. Alliance for Open Society International (2013). In other words, as long as the executive can sustain the claim that its AI procurement criteria are product specifications rather than viewpoint conditions, the order withstands challenge. This is precisely where the language of neutrality is weaponized to blur the line between science and ideology. The most significant feature that distinguishes EO 14319 from EO 13279 is the role that scientific and technical language plays in shielding ideological procurement behind plausible deniability. When the executive selectively excludes ideological adversaries in technical domains, the intervention can be reframed as quality assurance. The administration is not excluding DEI frameworks because it disagrees with their political implications; it is excluding them because federal agencies require "quality and accuracy of the output." The conditions sound epistemic rather than ideological, and epistemic criteria are the government's prerogative in procurement. Researchers at Stanford's Institute for Human-Centered AI have documented that genuine political neutrality in AI is practically unachievable. Data collections used to train AI models will always contain human inputs and embedded values (Stanford Institute for Human-Centered Artificial Intelligence, 2022). A truly "neutral" AI is, by definition, impossible. As a result, any administration can unilaterally declare an AI contractor's model to be biased — or neutral — without a verifiable standard to constrain that determination. The executive can control the procurement condition without statutory authorization or congressional oversight, hiding behind technical language that renders the ideological content of the criterion deniable. EO 14319 is an early instance of this pattern. The March 2026 executive order extending DEI prohibitions to all federal contracts and subcontractors at every tier, with False Claims Act enforcement, represents its acceleration. The trajectory is toward ideological governance through procurement criteria framed as quality standards, expanding in scope as the technical complexity of the regulated systems deepens the plausible deniability available to the executive. Congress cannot responsibly delegate to the executive the authority to define neutrality in technical domains unilaterally. When Congress delegates that authority, it delegates something that does not exist. There is no neutral ground in AI procurement ideology to which the executive can be directed. The delegation is therefore not a grant of authority to apply a standard; it is a grant of authority to invent one. An executive who invents the standard of neutrality it is required to enforce has been granted, in effect, the authority to define its own constitutional limits. That is not delegation — it is abdication. Congress is structurally unlikely to reclaim this authority because the punishment mode of procurement ideology is designed to avoid generating the conditions that would compel a congressional response. Bush's EO 13279 sustained ACLU lobbying campaigns, multiple congressional hearings, and years of constrained implementation precisely because the reward it offered was visible, its beneficiaries were identifiable, and its opponents were organized. EO 14319 produces none of these conditions. Its exclusions are concealed within product specifications and unverifiable deliberations. Its affected parties are AI vendors without a natural civil rights constituency. Its constitutional violations are insulated by the government speech doctrine and the inherent noise of competitive procurement. There is no organized opposition with a focused objective, because there is no organized political force with the political and legal standing to demand one. That gap between the constitutional obligation to constrain executive definition of technical neutrality and the political impossibility of meeting that obligation is the crisis this paper identifies. The Trump administration was successful in identifying and exploiting it. ________________
References
Agency for International Development v. Alliance for Open Society International, Inc., 570 U.S. 205 (2013). https://supreme.justia.com/cases/federal/us/570/205/ American Civil Liberties Union. (2003, March 25). ACLU statement on the unconstitutional or otherwise illegal conduct authorized by the proposed rule on faith-based organizations participation in HUD housing programs. https://www.aclu.org/documents/aclu-statement-unconstitutional-or-otherwise-illegal-conduct-authorized-proposed-rule-faith Goldenziel, J. (2004). The Bush faith-based initiative: Administratively quirky, constitutionally suspect. New York University Journal of Law and Public Policy. https://www.nyujlpp.org/wp-content/uploads/2012/11/Jill-Goldenziel-Administratively-Quirky.pdf Executive Order No. 13,279, 67 Fed. Reg. 77141 (Dec. 16, 2002). Equal protection of the laws for faith-based and community organizations. https://www.federalregister.gov/documents/2002/12/16/02-31831/equal-protection-of-the-laws-for-faith-based-and-community-organizations Executive Order No. 14,319, 90 Fed. Reg. 35389 (July 28, 2025). Preventing woke AI in the federal government. https://www.federalregister.gov/documents/2025/07/28/2025-14217/preventing-woke-ai-in-the-federal-government Goldenziel, J. (2004). The Bush faith-based initiative: Administratively quirky, constitutionally suspect. New York University Journal of Law and Public Policy. https://www.nyujlpp.org/wp-content/uploads/2012/11/Jill-Goldenziel-Administratively-Quirky.pdf Lawfare. (2025, August 7). Evaluating the "Woke AI" executive order. Lawfare Media. https://www.lawfaremedia.org/article/evaluating-the--woke-ai--executive-order National Institute of Family and Life Advocates v. Becerra, 585 U.S. __, 138 S. Ct. 2361 (2018). https://www.law.cornell.edu/supremecourt/text/16-1140 Personal Responsibility and Work Opportunity Reconciliation Act of 1996, Pub. L. No. 104-193, § 104, 110 Stat. 2105 (1996). https://www.congress.gov/104/plaws/publ193/PLAW-104publ193.pdf Reuters. (2026, April 21). Trump order targeting DEI by federal contractors challenged in lawsuit. Reuters. https://www.reuters.com/legal/government/trump-order-targeting-dei-by-federal-contractors-challenged-lawsuit-2026-04-21/ Rust v. Sullivan, 500 U.S. 173 (1991). https://supreme.justia.com/cases/federal/us/500/173/ Stanford Institute for Human-Centered Artificial Intelligence. (2022, June 28). AI's fairness problem: When treating everyone the same is the wrong approach. Stanford University. https://hai.stanford.edu/news/ais-fairness-problem-when-treating-everyone-the-same-is-the-wrong-approach