The Biden administration has announced a number of key updates to its AI executive order, including proposed rules for the training of powerful AI models and an AI talent hiring surge.
The executive order, which Biden signed in October, calls on the federal government to develop guidelines for the use of AI across the government, ensure the U.S. remains ahead of competing countries in AI development, and establish safeguards protecting against potential AI abuses including fraud, bias, and discrimination.
“The executive order lays out a number of activities across the board: innovation, regulation, civil rights, and on the 90 day timeframe, we’ve accomplished all of those actions,” explained White House Special Advisor for AI Ben Buchanan told Yahoo Finance Live on Tuesday.
According to a statement from the White House on Monday, the administration used the Defense Production Act to force the developers of the most powerful AI systems to report safety test results to the Commerce Department before they release those systems. Companies also have to tell the government when foreign clients use their platforms to train powerful AI models.
The idea is to ensure the government is aware of the kinds of models that are in development and how safe they are, and ensure rival nations do take advantage of those technologies.
The administration also kicked off an AI hiring surge to add more AI talent to government agencies, and started the EducateAI initiative to bring AI tools into the classroom. On Jan. 24, the National Science Foundation also announced its National AI Research Resource, which will provide community, rural, and minority college educators and researchers with access to high-powered AI resources.
“We have been overwhelmed by the talent that wants to come work for the government,” Buchanan said.
The executive order is the government’s most ambitious attempt to address the rapid growth of AI technology while trying to build out the U.S.’s AI capabilities. But critics have said it doesn’t address key issues like how AI could be used to improve citizen engagement.
Still others say an AI executive order or even Congressional legislation won’t be enough to deal with concerns related to AI misuse, and instead recommend the creation of a new government agency dedicated to the development and regulation of the technology.
Calls for AI regulation have increased over the past few months, spurred on by so-called deepfakes. Last week, AI-generative explicit images of Taylor Swift began appearing on the social media platform X. Prior to that a deepfake audio recording claiming to be President Biden was used in a series of phony calls to voters in New Hampshire. In 2023, an AI-generated image claiming to show an explosion outside of the Pentagon.
Members of Congress have been calling for some form of legislation seeking to govern AI, but nothing has come from those discussions.
According to Buchanan, watermarking, which embeds markings letting users know that an image or video is fake, could address some of those concerns. But it will take some time to get those rolled out, because the technology is still being developed.
“One of the things we did do…was the president secured voluntary commitments from the leading 15 AI companies in the United States last year to embed watermarks in their products…as they roll them out going forward,” Buchanan explained.
Still, if the government is going to regulate AI, Congress will need to step in at some point. And it’s difficult to say when something like that will actually happen.
Daniel Howley is the tech editor at Yahoo Finance. He’s been covering the tech industry since 2011. You can follow him on Twitter @DanielHowley.
