President Donald Trump signed an executive order Tuesday making a “voluntary framework” for AI corporations to share their frontier fashions with the federal authorities earlier than they’re launched “to advertise safe innovation and strengthen the cybersecurity of important infrastructure.”
The order says the US AI trade has succeeded partly “as a result of we refuse to stifle this innovation with overly burdensome regulation,” however that it additionally acknowledges new AI capabilities include safety dangers. Accordingly, it directs a number of federal companies to provide you with a framework to “assess the superior cyber capabilities of AI fashions” earlier than they’re launched to the general public. Corporations would have the discretion of whether or not to share their fashions with the federal government pre-release, however may get sure confidentiality protections in the event that they select to take action. It additionally requires the federal authorities to arrange cyber defenses for AI, particularly for important infrastructure.
The order comes after Trump postponed at the last minute signing a beforehand deliberate govt order that he worried may “get in the way in which” of competing with China. Whereas the sooner model would have allowed AI corporations to voluntarily share their fashions 14 to 90 days earlier than launch, based on The New York Timesthe present model asks corporations to share their fashions as much as 30 days earlier than public launch. Google, Microsoft, and xAI agreed last month to permit pre-release assessment by the Commerce Division’s Middle for AI Requirements and Innovation (CAISI).
OpenAI and Anthropic had already agreed to share their fashions with CAISI again in 2024 underneath President Joe Biden as a part of Biden’s push for AI safety guardrails. However till lately, the Trump administration downplayed security issues and took a hands-off method under former White House AI czar David Sacks. The order signed Tuesday explicitly says it shouldn’t be taken as a type of necessary licensing or preclearance. Nonetheless, it displays some willingness of the Trump administration to make use of oversight of AI corporations.
“The White Home is formally Mythos-pilled”
One issue on this shift could also be Anthropic’s restricted April rollout of its highly effective Mythos mannequin, which the corporate mentioned had flagged “hundreds of high-severity vulnerabilities, together with some in each main working system and internet browser.” Mythos additionally appeared to create an opening to thaw tensions between Anthropic and the administration, following its legal battle with the Pentagon over its use of AI for autonomous deadly weapons and mass surveillance.
The newly signed order has up to now garnered reward even from teams which have advocated towards restrictions on state AI legal guidelines. “The White Home is formally Mythos-pilled,” People for Accountable Innovation President Brad Carson mentioned in a press release, including that the order exhibits the Trump administration is taking AI vulnerabilities severely. Alliance for Safe AI CEO Brendan Steinhauser mentioned his group is “happy to see that the Trump administration is taking the dangers of those fashions severely,” and each Steinhauser and Carson urged Congress to codify necessary protections.
