A trustworthy source who has read the Biden Administration’s widely-anticipated AI executive order, which is poised to be released on Monday, told VentureBeat that the EO is “the longest” he could recall seeing, at over 100 pages.
Coincidentally, the discussion occurred last night at an AI ❤️ DC event hosted by VentureBeat and Anzu Partners with a rooftop view of the White House, during a week that included several AI-related gatherings in DC, including Wednesday’s second Senate AI Insight Forum.
There was plenty of chatter at the rooftop event about the upcoming AI executive order, which according to The Washington Post, would require “advanced AI models to undergo assessments before they can be used by federal workers.”
One attendee told VentureBeat she had been invited to a White House event scheduled for Monday afternoon that will accompany the release of the AI executive order. The Washington Post said the event, called “Safe, Secure, and Trustworthy Artificial Intelligence,” will be hosted by the president.
Another pointed out that the date of the EO release is no coincidence — coming just two days the UK’s AI Safety Summit — saying that the US clearly wants to show that they are leading on AI regulation. It also comes before EU officials reach a deal on the EU AI Act, which is anticipated to be passed by the end of the year.
The Use of Auditing and Assessments in AI Regulation
Several others told VentureBeat that the AI executive order, while limited, does use one of the Biden Administration’s few options to unilaterally tackle AI regulation — the federal government’s status and leverage as a top technology customer.
As the EO will require advanced AI models to undergo auditing and assessments before they can be used by federal workers, the idea is that this will encourage vendors selling to the federal government to make those audits part of their offering to their other non-government customers.
