Trump signs executive order on AI security: Voluntary 30-day review for frontier models contrasts with mandatory EU AI Act
What it really says
On June 2, 2026, President Trump signed the executive order 'Promoting Advanced Artificial Intelligence Innovation and Security.' It establishes a voluntary framework under which developers of frontier AI models can submit their systems to the US government for review up to 30 days before public release. The Treasury Department, the National Security Agency (NSA), and the Department of Homeland Security's Cybersecurity and Infrastructure Security Agency (CISA) are tasked with developing benchmarks to determine which models qualify as frontier AI. Additionally, an AI cybersecurity clearinghouse will be created that, in voluntary collaboration with industry, coordinates vulnerability scanning, validates discovered vulnerabilities, and prioritizes remediation and patch distribution. The executive order expressly states that it shall not be construed to authorize creation of any mandatory governmental licensing, pre-clearance, or permitting requirement for the development or release of AI models. An earlier draft had provided for a 90-day review period, which was shortened to 30 days over concerns about impeding innovation. The order exceeds in scope President Biden's October 2023 executive order, which had required developers of certain AI models to notify the government before training and share safety test results. However, Biden had not required developers to share their actual models with the government before market release. Trump had revoked Biden's AI order in January 2025 and initially pursued no regulation whatsoever.
Our assessment
This executive order marks a remarkable reversal: the same president who revoked Biden's AI safety order in January 2025 and spent 17 months without any regulation is now establishing his own review process for frontier models. This signals that even the world's most innovation-friendly government acknowledges that the risks of advanced AI systems cannot be ignored. At the same time, the limitations must be clearly stated: the framework is voluntary. No company is forced to submit its models to the government. There are no sanctions for non-participation and no public transparency requirements. Compared to the EU AI Act, which mandates risk classifications, documentation requirements, and penalties of up to 35 million euros, the American approach is significantly weaker. The cybersecurity clearinghouse is the most substantive part of the order, as it creates a concrete mechanism for coordinating vulnerability management. For consumers and citizens, the central question remains: Will the major AI companies voluntarily participate? Meta, Google, Anthropic, and OpenAI have not yet publicly committed to participation. Without their involvement, the framework remains a paper tiger.
Relevance for Germany
For Germany and Europe, this executive order is significant for several reasons. First, it highlights the fundamental difference in AI regulation philosophy between the US and EU: while the EU relies on mandatory rules with high fines through the AI Act, the US pursues a voluntary, incentive-based approach. For German companies operating in both markets, this means increasing regulatory complexity. Second, the order is an indirect compliment to the European approach: the fact that even the Trump administration acknowledges the need for some oversight after 17 months of laissez-faire supports the European position that frontier AI must be regulated. Third, the order has implications for transatlantic AI security cooperation: the cybersecurity clearinghouse could serve as a connecting point for cooperation with the BSI and ENISA to address AI system vulnerabilities across borders. Fourth, the contrast between the voluntary US framework and the mandatory EU AI Act, which becomes fully applicable on August 2, 2026, shows that Europe sets the world's strictest standards for AI regulation. This can be a competitive disadvantage but also a trust advantage for European AI products.
Fact check
The primary source is the full text of the executive order on whitehouse.gov along with the accompanying White House fact sheet. The 30-day voluntary submission period for frontier models is consistently reported by NPR, Scientific American, PBS News, Roll Call, and The Register. The involvement of Treasury, NSA, and CISA in benchmark development comes directly from the EO text and is confirmed by Federal News Network. The reduction from 90 to 30 days during the drafting process is reported by Scientific American and NPR citing government sources. The comparison with Biden's October 2023 EO and its revocation in January 2025 is documented through the official White House chronology. The characterization as a 'drastic reversal' comes from Scientific American, and the analysis of differences with the EU AI Act from the Council on Foreign Relations. The statement that the framework expressly does not create mandatory licensing is a direct quote from the executive order.
Source
- • https://www.whitehouse.gov/presidential-actions/2026/06/promoting-advanced-artificial-intelligence-innovation-and-security/
- • https://www.whitehouse.gov/fact-sheets/2026/06/fact-sheet-president-donald-j-trump-promotes-advanced-artificial-intelligence-innovation-and-security/
- • https://www.npr.org/2026/06/02/nx-s1-5844347/ai-safety-trump-executive-order
- • https://www.cfr.org/articles/assessing-trumps-executive-order-on-ai-oversight