In a move to balance national security with technological innovation, US President Donald Trump signed an executive order on Tuesday establishing a voluntary framework that allows the government to review advanced AI models before they are released to the general public.
Under the new provision, leading AI developers—including OpenAI, Google, and Anthropic—will grant the government access to their most powerful models for up to 30 days prior to their planned release. This window serves as a strategic compromise, following negotiations between the administration, which initially sought 90 days, and tech companies that pushed for a shorter 14-day period.
The order was prompted by security concerns surrounding Anthropic's "Mythos" model. The startup had previously withheld the model from public access due to its capability to identify vulnerabilities in critical systems, including those belonging to banks, hospitals, and government agencies.
Kent Walker, Google's head of public affairs, described the order as an "important step forward," stating that it ensures "defenders have the AI tools they need to keep America secure."
However, the road to the signing was not without tension. David Sacks, the venture capitalist serving as Trump's AI and crypto czar, had previously cautioned the president that such measures could hinder innovation. Sacks argued that unnecessary bureaucratic hurdles could weaken the United States' position in the competitive AI race with the Chinese mainland.
Beyond pre-release access, the executive order mandates the creation of an "AI cybersecurity clearinghouse." This initiative, involving the Treasury, the National Security Agency (NSA), and the CISA cybersecurity agency, will work voluntarily with industry leaders and critical infrastructure operators to identify and remediate software vulnerabilities.
This new framework replaces a Biden-era oversight order, which Trump scrapped upon returning to the White House. While the previous order focused on sharing safety test results, it was often criticized for being too lenient. In contrast, the European Union's AI Act, enacted in 2024, employs binding rules for high-risk systems, including mandatory transparency and incident reporting.
While some view the move as a positive step, critics argue that voluntary frameworks are insufficient. Anthony Aguirre, CEO of the Future of Life Institute, noted that the government must be empowered to block the release of systems that pose an "unacceptable national security risk" to ensure true safety.
Reference(s):
Trump signs AI order giving government access to powerful models
cgtn.com
