Will the US government take control of any AI company or project before 2030?

Explicit stances of specific individuals on DPA/IEEPA for AI tech are not detailed. The available research does not explicitly detail the official stance of specific individuals on Donald Trump's short-list for National Security Advisor (NSA) or Attorney General (AG) regarding the applicability of the Defense Production Act (DPA) or International Emergency Economic Powers Act (IEEPA) to compel private technology companies to hand over AI models or intellectual property during a national security crisis ^[1]. However, the broader context from available information indicates a strong focus on national security concerns related to advanced American AI technology, particularly in competition with China.

Potential Trump administration discussions consider DPA's relevance to AI firms. Within a potential future Trump administration, discussions have included scenarios where the DPA is considered relevant to AI firms. For instance, a hypothetical 2026 scenario depicts a future Trump administration directing federal agencies to cease using a prominent AI model, with the Pentagon identifying the startup as a supply risk, and the DPA highlighted as a potential tool in such situations ^[2]. The previous Trump administration also engaged in deals with AI chipmakers, indicating an interest in securing critical AI technology ^[4].

Some nominees express strong national security concerns regarding AI. While direct statements from individual short-listed candidates are not available, one "Trump CYBERCOM and NSA Nominee" has publicly warned about China's pursuit of advanced American AI chips ^[3]. This perspective signals a strong national security concern regarding critical AI technology, suggesting a potential openness to leveraging various governmental tools to safeguard such assets.

The U.S. AI Safety Institute actively models diverse catastrophic AI risks. Operating under the National Institute of Standards and Technology (NIST), the U.S. AI Safety Institute (AISI) is dedicated to identifying, evaluating, and mitigating 'catastrophic risk' scenarios associated with advanced AI models. Specific scenarios currently undergoing modeling include autonomous replication, cyberattack on critical infrastructure, and foreign adversary breakthroughs ^[1]. The AISI also prioritizes 'misuse risks' and 'dual-use' capabilities, which cover threats such as enabling widespread disinformation and fraud, assisting malicious cyber activities, creating biological or chemical weapons, and facilitating autonomous weapons systems ^[3]. Furthermore, the Institute aims to understand the dangers posed by self-improvement and deceptive capabilities of advanced AI systems, employing rigorous 'red-teaming' to test models for dangerous capabilities, particularly those related to chemical, biological, radiological, and nuclear (CBRN) materials and cybersecurity threats ^[2].

AISI's public plans omit federal seizure, though experts discuss possibilities. While current public documentation from the U.S. AI Safety Institute does not explicitly detail provisions for federal seizure or operational control of a responsible model or company, such measures are a topic of broader discussion among policy experts ^[2]. These discussions explore potential governmental actions to prevent or mitigate severe AI risks, suggesting that existing authorities, such as the Defense Production Act (DPA), could be invoked for national defense purposes. Invoking the DPA could address catastrophic harms from AI by enabling various forms of intervention or control ^[4]. However, proposals advocating for government seizure of AI platforms and data centers in specific scenarios are external analyses rather than direct evidence of such provisions within AISI's internal planning documents ^[5].

The publicly available agreement between Microsoft and OpenAI does not grant the government specific emergency control. The "Agreement between Microsoft Corporation and OpenAI Global, LLC" (EX-99.2) does not include explicit clauses for government 'step-in rights,' mandatory intellectual property transfer, or operational control during a declared national emergency ^[1]. While a force majeure clause (Section 10.3 "Governmental Actions") mentions "governmental actions" or "national or regional emergencies," this provision is intended to excuse contractual obligations between Microsoft and OpenAI, rather than to confer explicit control or facilitate asset transfer to a governmental entity ^[1].

Public Google-Anthropic documents similarly omit government emergency control clauses. Publicly available documents related to the partnership between Google and Anthropic, such as Anthropic's "Service Specific Terms," "Commercial Terms of Service," and End User License Agreement (EULA), are primarily customer-facing ^[2]. These documents require compliance with laws and allow Anthropic to respond to governmental requests for information. However, they do not contain specific clauses granting the government 'step-in rights,' intellectual property transfer, or operational control during an emergency within the scope of a strategic partnership ^[2]. A comprehensive strategic partnership agreement between Google and Anthropic that would detail such specific provisions is not present among the publicly accessible sources ^[3].

The Committee on Foreign Investment in the United States (CFIUS) holds extensive power to review foreign investments and impose safeguards. CFIUS possesses broad legal authority to review foreign investments in U.S. companies, assessing them for potential national security risks, and can impose stringent mitigation measures ^[3]. These measures may include requiring the divestiture of an investment, as evidenced by past presidential orders that mandated divestment from companies such as Jupiter Systems and Hiefo Corporation due to identified national security threats ^[4]. Beyond divestiture, CFIUS can also implement various mitigation agreements, which might involve independent auditors, monitoring agreements, or the appointment of a government-approved trustee or conservator to oversee a company's operations and ensure compliance with national security safeguards ^[1].

CFIUS reviews remain largely confidential, hindering specific disclosures on AI labs. The confidential nature of CFIUS reviews, with most transactions concluding without public disclosure, makes it inherently difficult to confirm any active, non-public reviews specifically targeting major AI labs with the potential for divestiture or conservatorship ^[1]. While major players like OpenAI and Anthropic are recognized for their strategic national security implications in the AI sector ^[2] and have been mentioned in contexts of government engagement, such as an OpenAI Pentagon deal or discussions surrounding an Anthropic supply-chain ban ^[5], none of the available sources explicitly confirm ongoing CFIUS reviews for these entities that could lead to a forced divestiture or conservatorship as a mitigation measure. Nonetheless, CFIUS's ongoing activity in reviewing major technology transactions is evident in public filings, illustrating the committee's vigilance in the tech sector ^[6].

The Trump campaign has drafted an executive order for federal AI preemption. This prepared contingency plan emphasizes establishing a uniform national artificial intelligence policy by preventing states from enacting their own potentially conflicting regulations ^[1]. Key technology advisors to the Trump campaign have been developing this approach, which seeks to streamline AI governance under federal authority ^[1]. Furthermore, Trump has publicly indicated his intention to repeal the existing Biden administration's AI executive order should he be elected ^[2].

The Biden campaign has not outlined new AI contingency plans. Available research does not detail specific new executive orders or alternative regulatory mechanisms currently being drafted by key technology advisors as a fallback should comprehensive AI safety legislation, such as Senator Schumer's SAFE Innovation Framework, fail to pass Congress by Q4 2025. While the current Biden administration has previously issued an executive order on artificial intelligence, the provided sources do not describe new contingency plans by the campaign's advisors for legislative gridlock extending beyond Q4 2025.