On January 20, 2025, President Donald Trump signed an executive order revoking Executive Order 14110, which had been issued by President Joe Biden on October 30, 2023. The revocation eliminated the federal AI safety framework that required developers of large AI models to share safety test results with the government and mandated risk management practices for AI systems. The action was part of a broader rollback of Biden-era technology regulations, signaling a shift toward less restrictive AI oversight in the United States.
The original order, titled "Safe, Secure, and Trustworthy Development and Use of Artificial Intelligence," had directed federal agencies to establish guidelines for AI safety, including stress-testing of models, watermarking of AI-generated content, and protections for privacy and civil rights. The revocation, announced on the first day of Trump's second term, was accompanied by a statement that the prior order "hinders AI innovation" and imposed "unnecessary burdens" on the private sector. The new administration emphasized a commitment to fostering American leadership in Artificial intelligence without what it characterized as excessive government interference.
Background: The 2023 Executive Order
Executive Order 14110 was signed on October 30, 2023, and represented the most comprehensive U.S. federal action on AI to that date. It invoked the Defense Production Act to require companies developing large AI models (those with significant computing power) to notify the Department of Commerce before releasing models to the public. The order also directed the National Institute of Standards and Technology (NIST) to create guidelines for AI safety, including red-team testing and evaluation of model robustness. Additionally, it mandated federal agencies to address AI-related risks in areas such as healthcare, education, and housing, and to develop standards for watermarking AI-generated content to help identify synthetic media.
The order was widely seen as a response to the rapid advancement of Generative AI systems, including Large language models from companies like OpenAI and Anthropic. It aimed to balance innovation with safeguards against potential harms, such as bias, disinformation, and national security threats. The order also established the AI Safety Institute within the Department of Commerce to conduct research and coordinate federal efforts.
The Revocation Order
The revocation was executed through a new executive order titled "Initial Rescissions of Harmful Executive Orders and Actions," signed on January 20, 2025. This order explicitly repealed Executive Order 14110 and directed all federal agencies to rescind any actions taken under it. The new order also revoked several other Biden-era policies, including those related to environmental and social justice, but the AI provisions were a focal point.
In a statement, the White House said that the previous order "impedes the development of AI" and "places the United States at a competitive disadvantage" relative to other nations, particularly China. The administration argued that AI regulation should be minimal and market-driven, allowing companies to innovate without federal mandates. The revocation was praised by many industry leaders, including executives from OpenAI, Anthropic, and Google DeepMind, who had previously expressed concerns about compliance costs and the potential for regulatory overreach.
Immediate Effects
The revocation immediately halted the implementation of several pending rules and guidelines. The Department of Commerce's AI Safety Institute, which had been established under the 2023 order, saw its mandate narrowed, though it was not disbanded. The requirement for companies to report safety test results to the government was lifted, and the NIST AI Risk Management Framework, while still available as a voluntary resource, was no longer a mandatory compliance standard for federal contractors.
Federal agencies were instructed to review and rescind any regulations or guidance issued in response to the 2023 order. This included the Office of Management and Budget's guidance on AI procurement and the Equal Employment Opportunity Commission's technical assistance on AI and workplace discrimination. The revocation also affected ongoing research initiatives, such as the National AI Research Resource, which had been piloted under the previous order, though its future remained uncertain.
Reactions and Criticism
The revocation drew sharp criticism from consumer advocacy groups, academic researchers, and some lawmakers. Critics argued that the 2023 order provided essential safeguards against AI-related harms, including algorithmic bias, privacy violations, and the spread of misinformation. They pointed to incidents of AI systems producing biased outcomes in hiring and lending, as well as the potential for AI to be used in disinformation campaigns. Some experts, such as Melanie Mitchell and Joshua Tenenbaum, expressed concern that the revocation would leave the U.S. without a coordinated federal response to AI risks, relying instead on voluntary industry self-regulation.
Several state governments, including California and New York, indicated they would continue to enforce their own AI regulations, which in some cases were stricter than the federal rules. However, the lack of a unified federal framework created a patchwork of laws that some companies said could complicate compliance.
Industry and Global Context
The revocation was part of a broader global debate on AI governance. The European Union had enacted the AI Act in 2024, which imposed binding requirements on AI systems based on their risk level. In contrast, the U.S. approach under the new administration favored a lighter touch, with an emphasis on voluntary standards and international competitiveness. The revocation was seen as a victory for tech companies that had lobbied against mandatory reporting and testing, arguing that such requirements would slow innovation and increase costs.
Some industry observers noted that the revocation might affect the development of AI in sectors such as healthcare and finance, where federal guidance had provided clarity on acceptable practices. For example, the Intuitive Surgical company, which uses AI in robotic surgery, had relied on the 2023 order's provisions for AI in medical devices. Without federal guidance, companies in these sectors might face greater uncertainty, though they could still seek approval from agencies like the Food and Drug Administration on a case-by-case basis.
Technical and Research Implications
The revocation also had implications for AI research and development. The 2023 order had funded research into AI safety, including work on Model Pruning, Data Augmentation, and Mechanistic interpretability. While these research areas continued, the loss of federal coordination meant that some projects lost their primary funding source. Universities such as Stanford AI Lab and BAIR (Berkeley AI Research) had received grants under the order; these grants were not immediately canceled, but future funding was uncertain.
In the absence of federal mandates, many AI companies announced they would continue to follow voluntary safety practices, such as Red teaming (AI) and Reinforcement Learning from AI Feedback (RLAIF) (reinforcement learning from AI feedback). However, the lack of enforcement raised concerns among researchers like Aleksander Madry and Anima Anandkumar that safety standards might be unevenly applied across the industry.
Legal and Political Challenges
The revocation was immediately challenged in court by several advocacy groups, who argued that the president lacked the authority to rescind an executive order without following the Administrative Procedure Act. Legal experts noted that executive orders are not laws and can be revoked by a subsequent president, but the challenge centered on the fact that the 2023 order had been used to implement statutory requirements, such as those under the Defense Production Act. As of early 2025, the legal status of the revocation remained unresolved, with hearings scheduled in federal court.
Politically, the revocation became a flashpoint in the debate over AI regulation. Some Republican lawmakers praised the move as a necessary step to boost American competitiveness, while Democrats and some independents called for new legislation to codify AI safety requirements. The issue was expected to be a major topic in the 2026 midterm elections.
Future Outlook
Following the revocation, the future of U.S. AI policy remained uncertain. The administration indicated that it would work with industry to develop voluntary standards, and the Department of Commerce was tasked with facilitating private-sector initiatives. However, no concrete plan had been announced as of February 2025. Some experts predicted that the U.S. would fall behind in AI safety research, while others argued that innovation would accelerate without regulatory burdens.
The revocation also had international repercussions. Other countries, including the United Kingdom and Japan, were watching the U.S. approach closely as they formulated their own AI policies. The shift in U.S. policy was seen as a potential obstacle to international cooperation on AI governance, such as the Bletchley Declaration signed in 2023, which had called for a global consensus on AI risks.
In the meantime, companies like OpenAI, Anthropic, and Google DeepMind continued to develop advanced AI systems, but without federal oversight, the responsibility for ensuring safety fell increasingly on the private sector. Whether this approach would prove sustainable remained an open question, with many stakeholders calling for a balanced framework that would protect the public without stifling innovation.