Trump signs AI safety executive order: Voluntary 30-day model review before release - originally planned 90-day window shrunk after lobbying by Musk and Zuckerberg
What it really says
On June 2, 2026, President Trump signed the executive order 'Promoting Advanced Artificial Intelligence Innovation and Security.' The order establishes a voluntary framework under which AI companies can submit their most powerful models to the government for security testing up to 30 days before public release. The Treasury Department, the National Security Agency, and the Cybersecurity and Infrastructure Security Agency (CISA) are tasked with developing benchmarks to define which models qualify as frontier AI. Within 30 days of signing, an AI cybersecurity clearinghouse will be established in voluntary collaboration with industry to coordinate the detection, validation, and remediation of software vulnerabilities. CISA receives authority to issue binding operational directives to protect civilian federal agencies and to extend AI-powered defensive tools to municipalities, rural hospitals, and critical infrastructure operators. The order explicitly states that it does not create mandatory licensing, pre-clearance, or permitting requirements for new AI models.
Our assessment
This executive order deserves a nuanced evaluation. On the positive side, the US government is creating a structured framework for AI model security testing and establishing a centralized clearinghouse for cybersecurity vulnerabilities. However, the purely voluntary nature is a significant limitation: no AI company is required to submit its models for pre-release testing. The order's history reveals massive industry influence: the originally planned 90-day review window was cut to 30 days after direct intervention by Elon Musk, Mark Zuckerberg, and venture capitalist David Sacks. The signing ceremony was initially scheduled for May 21 but was postponed after industry protests. Trump's justification: he did not want to slow American companies in the competition with China. The contrast with the European approach could hardly be starker. While the EU AI Act establishes mandatory requirements backed by fines of up to 35 million euros or 7 percent of global annual revenue, the US relies on voluntary cooperation. For global AI safety, this means the regulatory patchwork continues to grow, and the world's largest AI market remains without mandatory pre-release testing requirements.
Relevance for Germany
For Germany, this executive order has several implications. First, it underscores the fundamentally different regulatory approaches: while Germany is advancing national implementation of the binding EU AI Act through the AI Market Surveillance and Innovation Promotion Act, the US approach remains voluntary. German companies using AI models from American providers like OpenAI, Anthropic, or Google face a paradoxical situation: these models undergo no mandatory pre-release testing in the US but must meet strict requirements in the EU. Second, the cybersecurity clearinghouse also affects European interests: vulnerabilities in AI systems operated from US data centers can directly endanger European users and critical infrastructure. Whether and how findings will be shared with European partners like ENISA is left open by the order. Third, the process confirms many European politicians' concerns that US regulation is primarily driven by industry interests: direct intervention by tech billionaires measurably weakened the order. This is an important signal for upcoming discussions in the EU-US Trade and Technology Council.
Fact check
The primary source is the official executive order text on whitehouse.gov. Core facts are consistently confirmed by NPR, CNBC, PBS, Axios, and The Hill: voluntary 30-day review window, AI cybersecurity clearinghouse led by the Treasury Department, no mandatory licensing requirements. Information about the reduction from 90 to 30 days and the intervention by Musk, Zuckerberg, and Sacks comes from Axios citing an informed source and from TechTimes. The originally planned May 21 signing and subsequent postponement after industry protests is independently confirmed by NPR, Axios, and the Washington Post. The role of David Sacks and National Economic Council deputy director Ryan Baasch in weakening the provisions was reported by Axios.
Source
- • https://www.whitehouse.gov/presidential-actions/2026/06/promoting-advanced-artificial-intelligence-innovation-and-security/
- • https://www.npr.org/2026/06/02/nx-s1-5844347/ai-safety-trump-executive-order
- • https://www.cnbc.com/2026/06/02/trump-executive-order-ai.html
- • https://www.axios.com/2026/06/02/trump-signs-new-ai-executive-order