Trump pulls AI security executive order at last minute: Zuckerberg and Musk intervened
What it really says
On Thursday, May 21, 2026, President Donald Trump pulled a sweeping AI security executive order just hours before the planned signing ceremony. The order would have established a voluntary framework requiring AI companies to share frontier models with the government at least 90 days before public release for security review. The NSA would have played a central role in evaluating dangerous capabilities and security vulnerabilities. Additional cybersecurity measures were included: protection for critical infrastructure such as hospitals and banks, expanded cyber hiring at the Pentagon, and a clearinghouse at the Treasury Department for AI security vulnerabilities. Trump told reporters in the Oval Office: 'I didn't like certain aspects of it' and 'We're leading China, we're leading everybody, and I don't want to do anything that's going to get in the way of that lead.' Behind the scenes: between Wednesday evening and Thursday morning, both Meta CEO Mark Zuckerberg and xAI CEO Elon Musk spoke with Trump by phone. AI adviser David Sacks also opposed the order. According to Axios sources, the primary reason was that Trump 'just hates regulation.' The White House stated it is 'studying' the order further, with no new signing date announced.
Our assessment
This development is serious. The shelved order would have been the first meaningful US regulatory instrument for AI safety after Trump revoked Biden's 2023 AI executive order in January 2025. Even this voluntary, industry-friendly framework - not a mandate, but a 90-day pre-release review on a voluntary basis - was apparently too much for tech CEOs. The intervention by Zuckerberg and Musk demonstrates how directly the AI industry influences policy. The result: the world's largest AI companies - OpenAI, Google, Meta, Anthropic, xAI - are all US-based and now operate without any federal AI safety framework. Frontier models with potentially dangerous capabilities in cybersecurity, biological weapons, or autonomous agents can be released without any government security review. For global AI safety, this is a setback: the EU AI Act can regulate the deployment of AI in Europe, but cannot control the development and initial release of models built in the United States.
Relevance for Germany
For Germany and the EU, this decision significantly deepens the transatlantic regulatory gap. While the EU introduces increasingly strict rules through the AI Act and the Digital Omnibus agreed on May 7, the US is moving in the opposite direction - even voluntary safety reviews are being rejected. German companies using AI products from US providers face a dilemma: they must comply with the EU AI Act while using models that have undergone no safety review in the US. Germany's BSI, which will serve as the national market surveillance authority for the AI Act from September 2026, will particularly feel the consequences of this regulatory gap. The planned 90-day pre-release review would have been the first US mechanism roughly comparable to the EU AI Act's conformity assessments. Its failure strengthens the argument of those in the EU calling for greater digital sovereignty and European AI infrastructure, including voices from Germany's Federal Ministry for Economic Affairs and the European Commission.
Fact check
The postponement of the executive order is independently confirmed by at least six major US news organizations: CNBC, Bloomberg, Washington Post, Axios, NBC News, and CNN. Trump's quotes come from White House pool reports from May 21, 2026. The intervention by Zuckerberg and Musk is reported by Axios and NBC News citing informed sources. The content of the planned order - voluntary 90-day framework, NSA involvement, cybersecurity measures - is consistently described. David Sacks' role as an opponent of the order comes from Axios sources. The White House has confirmed the postponement without naming a new date. Background: Trump revoked Biden's Executive Order 14110 on AI safety on January 20, 2025. The now-shelved draft was the first attempt at Trump-era AI regulation.
Source
- • CNBC 21.05.2026: Trump postpones AI executive order signing (cnbc.com/2026/05/21/trump-ai-executive-order-postponed.html)
- • Axios 21.05.2026: Why Trump's AI executive order was pulled (axios.com/2026/05/21/trump-ai-executive-order-postponed-why)
- • Washington Post 21.05.2026: Trump calls off plan to sign artificial intelligence order (washingtonpost.com/business/2026/05/21/trump-ai-executive-order/)
- • TechCrunch 21.05.2026: Trump delays AI security executive order (techcrunch.com/2026/05/21/trump-delays-ai-security-executive-order-i-dont-want-to-get-in-the-way-of-that-leading/)
- • NBC News 21.05.2026: Trump abruptly scraps signing of landmark executive order regulating AI (nbcnews.com/tech/tech-news/trump-scraps-signing-landmark-executive-order-regulating-ai-rcna346288)
- • Berliner Zeitung 05.2026: Trump administration plans U-turn on AI oversight (berliner-zeitung.de)