Trump Order: US AI Firms to Voluntarily Test Models for Cybersecurity

On June 2, 2026, then-President Donald Trump signed an executive order titled 'Promoting Advanced Artificial Intelligence Innovation and Security,' which has significant implications for the development and deployment of cutting-edge AI models in the United States. Contrary to the initial sensationalized headline from The Times of India, which suggested a 'banning' of free AI model release, the executive order actually establishes a *voluntary* framework for pre-release cybersecurity testing of powerful AI systems. The core provision of the order encourages leading American artificial intelligence companies, such as OpenAI, Google, and Anthropic, to voluntarily provide the US government with access to their most capable AI models for cybersecurity assessments. This access would be granted for a period of up to 30 days before these models are released to other trusted partners or the general public. The objective is to allow federal cybersecurity experts to rigorously test these advanced AI systems for potential digital vulnerabilities and safety flaws, thereby strengthening national security and protecting critical infrastructure. This executive order represents a notable shift in the Trump administration's approach to AI regulation, which had previously leaned towards a more deregulatory and hands-off stance to foster innovation. However, escalating national security concerns in Washington, particularly regarding the potential for ultra-advanced AI systems to identify and exploit cybersecurity weaknesses, prompted this move towards more active government oversight. The order specifically referenced concerns over Anthropic's 'Mythos' model, which reportedly demonstrated an ability to uncover vulnerabilities in various computer systems, including those of banks, hospitals, and government entities. Multiple federal departments and agencies are tasked with implementing this framework, including the Department of the Treasury, Department of Defense (specifically the National Security Agency), Department of Commerce, and Department of Homeland Security (through the Cybersecurity and Infrastructure Security Agency or CISA). These agencies are directed to collaborate with the AI industry to establish voluntary testing agreements. Additionally, the order mandates the formation of an 'AI cybersecurity clearinghouse' to coordinate vulnerability scanning, validation, and remediation efforts in voluntary collaboration with the AI industry and critical infrastructure operators. Initially, an earlier draft of the executive order proposed a 90-day review period for AI models, but this was shortened to 30 days following concerns that a longer window could impede American competitiveness against countries like China in the global AI race. The White House clarified that this process is explicitly voluntary and does not constitute a mandatory licensing or pre-clearance regime, emphasizing that excessive government overreach could stifle innovation and free speech. This voluntary nature has been welcomed by some industry leaders, with OpenAI CEO Sam Altman stating that the order strikes the right balance between developing advanced models and ensuring their safety. Beyond cybersecurity, the Trump administration has a broader history of engaging with AI policy. Earlier initiatives during his presidency (2017-2021) included Executive Order 13859 in February 2019, which established the American AI Initiative to promote leadership in AI research and development. Subsequent orders in July and December 2025 and even earlier in January 2025, focused on accelerating data center construction, promoting US AI technology exports, addressing perceived biases in federal AI procurement, and establishing a national policy framework to potentially override state-level AI regulations. These actions underscore a consistent effort to secure American AI leadership, often framed in the context of global competition, particularly with China. In summary, the executive order of June 2, 2026, does not ban the free release of AI models but instead institutes a voluntary government review process for advanced AI systems to address burgeoning national security risks. This nuanced approach seeks to balance the promotion of AI innovation with the imperative of safeguarding critical infrastructure and national interests in an increasingly AI-driven world.

Frequently Asked Questions

Read Full Story on Quick Digest