概述
The main ones are EO 13859 (2019), EO 14110 (2023), and EO 14179 (2025), which followed the revocation of EO 14110. They matter because, with no comprehensive federal AI law, these orders have been the main way Washington sets national AI policy, and each new administration can undo the last one's.
深入探讨
An executive order is not a statute. It directs the executive branch, and it can reach private companies only when it rests on authority Congress already granted. It also lasts only as long as a President keeps it in place. That explains how US AI policy has swung back and forth. EO 13859, 'Maintaining American Leadership in Artificial Intelligence,' was signed in February 2019 during President Trump's first term and launched the American AI Initiative. It told agencies to prioritize AI research funding, make federal data and computing more available to researchers, and train the workforce. It also directed OMB to write guidance on regulating AI and NIST to plan technical standards. EO 13960, signed in December 2020, set principles for trustworthy government use of AI and started agency use-case inventories. EO 14110, signed by President Biden on October 30, 2023, was far broader. It invoked the Defense Production Act to require developers of the most powerful dual-use foundation models to report safety test (red-team) results and training details to the Commerce Department. It asked NIST to develop evaluation guidance, and NIST created the US AI Safety Institute, later renamed the Center for AI Standards and Innovation. It also pushed guidance on labeling synthetic content and told OMB to set governance rules for agencies, which became memo M-24-10. On January 20, 2025, President Trump revoked EO 14110. Days later EO 14179, 'Removing Barriers to American Leadership in Artificial Intelligence,' ordered an AI Action Plan. OMB replaced its guidance in April 2025. America's AI Action Plan was released in July 2025 alongside orders on data center permitting, AI exports and 'unbiased' AI in federal procurement. A common misconception is that revoking EO 14110 wiped out all federal AI governance. Agency inventories, Chief AI Officers and NIST's work continued in revised forms.
战略影响
风险与安全
灾难性和日常的人工智能危害都取决于谁了解风险以及谁能够采取行动。
更清晰的判决
公众和专业素养决定强有力的安全政策在政治上是否可行。
打破炒作
清晰的解释可以减少炒作、实验室公关和模糊道德剧场的影响。
The Future of US Presidential Executive Orders on AI
Federal AI policy will likely keep shifting with administrations until Congress passes durable legislation, and no comprehensive AI statute has passed so far. Expect continued use of procurement rules, export controls and infrastructure policy as the main levers, since those rest on existing authority. The relationship between federal action and state AI laws is now a live legal question, and courts may define how far executive preemption efforts can reach. Organizations should treat executive orders as signals of direction and watch the implementing guidance, which is where concrete obligations appear.
现实世界的实施
A federal agency names a Chief AI Officer and publishes an inventory of its AI use cases, as White House and OMB guidance under these orders has required.
In 2024, companies training frontier models above a very large compute threshold faced reporting obligations to the Commerce Department under EO 14110, which relied on the Defense Production Act.
A procurement officer at a federal agency rewrites contract language for a chatbot purchase after new OMB memos in 2025 replaced the Biden-era guidance on buying AI.
A state legislator follows a late-2025 executive order directing federal action against state AI laws the administration considers burdensome, and weighs whether their own state's bill could be challenged.
风险与防护栏
将存在风险视为科幻小说，同时能力复合。
混淆了表面产品安全与高度自治下的对准。
只给非英语和非专业观众留下低质量的资源。
实施路线图
单独的产品危害、误用和失控/失调风险。
询问哪些证据会改变您对时间表和严重性的看法。
比起营销主张，更喜欢主要来源和具体评估。
确定一条行动路径：职业、政策、资金或技能——而不仅仅是意识。
不断探索
Free newsletter
Get the daily AI briefing
Three verified AI stories every weekday morning, written in plain English. Free forever, no ads.
One email each weekday. Unsubscribe in one click. We never sell or share your address.
Test yourself
Take the US Presidential Executive Orders on AI quiz
Instant feedback on every answer, and a shareable certificate with a verifiable ID once you pass a course.
Support free AI education. AI Understanding is a 501(c)(3) nonprofit — no ads, no paywall, ever. Make a donation
常见问题
What is US Presidential Executive Orders on AI?
US presidential executive orders on AI are directives from the President telling federal agencies how to research, use, buy and oversee artificial intelligence. The main ones are EO 13859 (2019), EO 14110 (2023), and EO 14179 (2025), which followed the revocation of EO 14110. They matter because, with no comprehensive federal AI law, these orders have been the main way Washington sets national AI policy, and each new administration can undo the last one's.
Which executive order launched the American AI Initiative in 2019?
EO 13859, 'Maintaining American Leadership in Artificial Intelligence,' was signed in February 2019 and launched the American AI Initiative focused on research, data access, standards and workforce.
What legal authority did EO 14110 rely on to require frontier model developers to report safety test results?
EO 14110 invoked the Defense Production Act, which let it compel information from private companies rather than only directing agencies.
How was the reporting threshold for the most powerful models defined under EO 14110?
The order used training compute as the trigger because it can be measured before capabilities are known, with a lower threshold for models trained mainly on biological sequence data.
What happened to EO 14110 on January 20, 2025?
President Trump revoked EO 14110 on his first day back in office, and EO 14179 soon ordered a new AI Action Plan.
What did EO 14179 direct the administration to produce?
EO 14179, 'Removing Barriers to American Leadership in Artificial Intelligence,' ordered an AI Action Plan, which was released in July 2025.
继续学习
相关指南
为此主题精选的更多指南