President Donald Trump on Tuesday signed an executive order generating a voluntary framework under which AI developers will share advanced models with the government before public launch.
The central provision permits companies which includes OpenAI, Google and Anthropic to give the authorities to their most powerful models for as much as 30 days before deliberate launch.
The order was caused via concerns over Anthropic’s Mythos model, which the AI startup has held again from the public due to its ability to expose vulnerabilities in computer systems, inclusive those of banks, governments and hospitals.
The 30-day window signifies a compromise. The original draft called for up to 90 days of pre-launch authorities access, while tech companies had driven to reduce that figure to just 14 days.
For OpenAI chief Sam Altman, the executive order “get the balance proper.”
“The U.S. should lead on AI by persisting to create the very best models, making sure they’re safe, and getting cyber tools into the hands of trusted defenders,” Altman stated.
Kent Walker, Google’s head of public affairs, referred to as the order an “crucial breakthrough” that will ensure “defenders have the AI tools they want to keep America secure.”
Anthropic, which has regularly clashed with the Trump administration, called the order “an vital step in reinforcing America’s leadership in AI.”
Unnecessary
The signing comes after a turbulent few weeks in which the White House emerged close to unveiling the measure, only to drag again abruptly.
As per Politico and other media, David Sacks, the Silicon Valley project capitalist who served as Trump’s AI and crypto czar, called the president to warn that the measure might sluggish innovation and hurt the US in its AI race with China—blindsiding White House staff members who believed Sacks supported the order.
Sacks wrote on X last week that “needless regulation is the biggest risk to innovation in America,” including that winning the AI race needed clearing “bureaucratic obstacles” from state legislatures and “woke” Washington politicians.
The order also guides the Treasury Department, the National Security Agency and the Cybersecurity and Infrastructure Security Agency to form an “AI cybersecurity clearinghouse” in voluntary collaboration with industry and vital infrastructure operators to detect software vulnerabilities and find approaches to fix them.
Trump scrapped an AI oversight order from his Democratic predecessor, Joe Biden, on his first day back in the White House.
Biden’s 2023 order needed AI companies to share safety test outcomes with the government and depends closely on voluntary commitments—already a light-touch technique that fell short of what many professionals had referred for.
By assessment, the European Union’s AI Act—which entered into force in 2024—units binding rules for high-risk AI systems, along with mandatory transparency necessities and, for the most powerful models, responsibilities around safety testing out and incident reporting.
“This is an essential step in the right direction,” stated Anthony Aguirre, CEO of the Future of Life Institute, which advocates for AI safety.
“Voluntary frameworks are not enough, but.”
The government have to be empowered “to block the launch of systems that pose an unacceptable national safety chance,” he added.
