In a move to strengthen national security and safeguard critical infrastructure, US President Donald Trump signed an executive order this past Tuesday, establishing a voluntary framework that requires AI developers to share their most advanced models with the government prior to public release.
The core of the order allows industry leaders such as OpenAI, Google, and Anthropic to provide the government with access to their most powerful models for a window of up to 30 days before they are made available to the general public. This measure was largely prompted by concerns surrounding Anthropic's Mythos model, which the startup has withheld from the public due to its potential to expose critical vulnerabilities in computer systems used by hospitals, governments, and financial institutions.
The 30-day period is the result of a strategic compromise. While the original draft of the order proposed a 90-day government review period, tech companies had lobbied for a shorter window of just 14 days. Kent Walker, Google's head of public affairs, described the order as an "important step forward," emphasizing that it ensures defenders possess the necessary AI tools to maintain security.
However, the road to the signing was not without internal friction. Reports indicate that David Sacks, the venture capitalist serving as the president's AI and crypto czar, warned that such measures could stifle innovation. Sacks argued that unnecessary regulation could hinder the United States in its competitive AI race with the Chinese mainland, stressing the need to remove bureaucratic hurdles to maintain a technological edge.
Beyond the pre-release access, the executive order mandates the creation of an "AI cybersecurity clearinghouse." This entity, formed by the Treasury, the National Security Agency (NSA), and the CISA cybersecurity agency, will work in voluntary collaboration with industry partners to identify and remediate software vulnerabilities.
This new approach marks a significant shift in policy. Upon returning to the White House, President Trump scrapped a Biden-era AI oversight order from 2023, which had relied on safety test results and voluntary commitments. In contrast to the US approach, the European Union's AI Act, which took effect in 2024, imposes binding rules and mandatory transparency for high-risk AI systems.
While some see the new order as a victory for security, some advocates argue it does not go far enough. Anthony Aguirre, CEO of the Future of Life Institute, noted that while it is a step in the right direction, voluntary frameworks are insufficient. He argued that the government must have the power to block the release of systems that pose an unacceptable risk to national security.
Reference(s):
Trump signs AI order giving government access to powerful models
cgtn.com
