White House to issue an executive order on govt review of advanced AI models before launch
Washington DC, May 21
The White House is expected to issue an executive order as early as Thursday proposing a voluntary framework for government review of advanced artificial intelligence models before their public release, CNN reported, citing sources familiar with the matter.
Under the proposed plan, AI companies would share their frontier models with US authorities for evaluation ahead of launch. The timeline for such pre-release access remains under discussion, with one draft suggesting a 90-day review window, while some industry players are reportedly pushing for a shorter period of around 14 days.
Major AI firms, including OpenAI and Anthropic, are said to be engaged in discussions with the White House over the proposed framework. The initiative is aimed at assessing potential risks posed by advanced AI systems, particularly in areas such as cybersecurity, where experts warn that powerful models could be misused to accelerate cyberattacks, CNN reported.
According to CNN's sources, the draft executive order is divided into two key sections, cybersecurity and "covered frontier models." The latter outlines which AI systems would qualify for the voluntary pre-release review mechanism. The cybersecurity section reportedly proposes the creation of a voluntary "clearinghouse" involving the Treasury Department, other federal agencies and private AI firms to identify and address vulnerabilities in unreleased models.
The proposal also includes plans to expand hiring under the US Tech Force, a group of engineers tasked with modernising government IT systems.
The Trump administration had previously adopted a relatively hands-off approach to AI regulation, though recent developments, including Anthropic's unreleased Mythos model and its controlled access programme, have prompted closer engagement between the government and industry on safety concerns.
Meanwhile, the Commerce Department's National Institute of Standards and Technology had earlier indicated that major tech companies would share unreleased AI models for national security evaluations, although the announcement has since been removed from the department's website.
The White House has not issued any official comment on the reported executive order.
— ANI
Reader Comments
Voluntary framework? That's just a polite suggestion, not regulation. 😒 These companies have too much power already. If they really cared about safety, they'd make it mandatory.
Interesting how the US is now taking AI risks seriously after that Mythos model leak. But 90 days review? That's too long! In India, we need faster clearance for innovation, but also safety checks. Balancing act, yaar.
The cybersecurity angle is smart. AI can be weaponized for attacks, so having the Treasury and other agencies involved makes sense. But will this stifle innovation? That's the real question.
Voluntary review means nothing if companies can just ignore it. We need binding rules, especially for AI that affects elections, jobs, and national security. India should watch this closely and learn from their mistakes.
Finally some action! But why voluntary? 🤔 If AI can cause widespread cyberattacks, that's a national security threat. Should be mandatory. Also, 14 days vs 90 days debate shows industry is resisting real oversight. Typical.
As an AI researcher in Bangalore, I think this is a good start. But the devil is in the details - which models qualify and who reviews
We welcome thoughtful discussions from our readers. Please keep comments respectful and on-topic.