US President Donald Trump abruptly canceled the signing of a proposed artificial intelligence executive order that would have mandated government evaluation of powerful AI models before public release. The decision came hours before a scheduled Oval Office event, following reports that several top technology executives could not attend on short notice. Trump stated that he delayed the policy to avoid impeding America’s competitive global edge, noting, “I think it gets in the way of – you know, we’re leading China, we’re leading everybody.” The administration had invited CEOs from OpenAI, Google, Anthropic, Meta, and Microsoft just 24 hours prior.

The regulatory initiative was originally prompted by Anthropic's restricted model, Mythos. This system specializes in autonomously identifying complex software vulnerabilities at scale, having achieved an 83.1% success rate in internal tests for reproducing known security flaws. The model reportedly identified thousands of "zero-day" vulnerabilities across major browsers and operating systems, including a 27-year-old security flaw in OpenBSD. According to Logan Graham, Anthropic’s red team lead, Mythos can systematically discover tens of thousands of flaws and independently write exploit code, a significant increase from previous models like Claude Opus 4.6, which identified approximately 500 such issues. The National Security Agency (NSA) reportedly began utilizing the model to probe weaknesses in US government systems, leading Anthropic to restrict access to roughly 40 handpicked cybersecurity and technology firms.

The canceled executive order would have required tech companies to submit their most advanced systems to the Office of the National Cyber Director (ONCD) for months-long pre-release evaluations. Industry leaders, including Google, Microsoft, Meta, and OpenAI, faced potential compliance bottlenecks and product rollout delays under this framework. By abandoning the mandate, the administration allows these companies to continue deploying and monetizing their rapidly evolving systems without waiting for government approval. The reversal marks a significant pivot from the initial security concerns triggered by the capabilities of the Mythos model, reflecting a broader administration focus on maintaining the US technological lead.

US President Donald Trump abruptly canceled the signing of a proposed artificial intelligence executive order that would have mandated government evaluation of powerful AI models before public release. The decision came hours before a scheduled Oval Office event, following reports that several top technology executives could not attend on short notice. Trump stated that he delayed the policy to avoid impeding America’s competitive global edge, noting, “I think it gets in the way of – you know, we’re leading China, we’re leading everybody.” The administration had invited CEOs from OpenAI, Google, Anthropic, Meta, and Microsoft just 24 hours prior.

The regulatory initiative was originally prompted by Anthropic's restricted model, Mythos. This system specializes in autonomously identifying complex software vulnerabilities at scale, having achieved an 83.1% success rate in internal tests for reproducing known security flaws. The model reportedly identified thousands of "zero-day" vulnerabilities across major browsers and operating systems, including a 27-year-old security flaw in OpenBSD. According to Logan Graham, Anthropic’s red team lead, Mythos can systematically discover tens of thousands of flaws and independently write exploit code, a significant increase from previous models like Claude Opus 4.6, which identified approximately 500 such issues. The National Security Agency (NSA) reportedly began utilizing the model to probe weaknesses in US government systems, leading Anthropic to restrict access to roughly 40 handpicked cybersecurity and technology firms.

The canceled executive order would have required tech companies to submit their most advanced systems to the Office of the National Cyber Director (ONCD) for months-long pre-release evaluations. Industry leaders, including Google, Microsoft, Meta, and OpenAI, faced potential compliance bottlenecks and product rollout delays under this framework. By abandoning the mandate, the administration allows these companies to continue deploying and monetizing their rapidly evolving systems without waiting for government approval. The reversal marks a significant pivot from the initial security concerns triggered by the capabilities of the Mythos model, reflecting a broader administration focus on maintaining the US technological lead.