US President Donald Trump has recently delayed signing a highly anticipated executive order on AI safety. The order was originally intended to require US government agencies to conduct security assessments of advanced AI models before their official release, but Trump stated that certain provisions could hinder America's leadership in artificial intelligence. In an interview with White House reporters, Trump expressed dissatisfaction with parts of the executive order, emphasizing: "We are leading China and everyone else, and I do not want to take any action that would undermine this lead." He further noted that some language in the document "could become an obstacle." According to multiple media reports, another informal reason for delaying the signature ceremony was that several tech company CEOs were unable to reach Washington within a short timeframe. Since the White House typically invites industry executives to attend such events, these logistical arrangements have also been cited as influencing the schedule. Previously disclosed details indicated that the executive order would authorize bodies including the Office of the National Director of Cyberspace to establish a security review mechanism specifically targeting AI models, focusing primarily on assessing potential cybersecurity risks associated with them. These discussions partly stem from external concerns regarding Anthropic's newly unveiled Mythos and OpenAI's GPT-5.5 Cyber among other emerging models believed capable of rapidly identifying and exploiting vulnerabilities—potentially introducing novel cyberattack threats. Additionally, one contentious clause in the draft required AI companies to share access to advanced models with the government between 14 and 90 days prior to public launch to facilitate security evaluations. Some industry insiders argue this requirement may increase burdens on corporate R&D and deployment processes while raising issues related to trade secret exposure.