Trump promised to repeal Biden’s AI executive order — here’s what to expect next

President-elect Donald Trump has vowed to repeal a Biden administration executive order on artificial intelligence intended to erect guardrails around the technology in the absence of congressional action. 

What happens next for the associated policies governing how federal agencies use AI isn’t totally clear, although anti-bias provisions in existing guidance may be on the chopping block.

The 2024 Republican platform describes Biden’s executive order as “dangerous,” saying that it “hinders AI Innovation, and imposes radical leftwing ideas on the development of this technology.”

At a rally late last year, Trump said he would ax the order on day one.

“Republicans support AI development rooted in free speech and human flourishing,” the Republican platform states. The Trump campaign did not respond to requests for comment for this story.

Among the policies flowing from the order, issued last fall, are safeguard mandates for federal agencies around the use of AI and guidance on how the federal government purchases the technology. 

Last month, the Biden administration also unveiled the first-ever national security memorandum on the use of AI and an associated framework for intelligence agencies and national security institutions.

The implementation memo for the executive order includes a raft of requirements meant to help agencies manage the risks associated with AI when it could impact people's safety or rights, like when it’s used at the border. 

Under the guidance, AI applications deemed to be rights-impacting have to be assessed in terms of their impact on equity and fairness. 

“If the [Department of Veterans Affairs] wants to use AI in VA hospitals to help doctors diagnose patients, they would first have to demonstrate that AI does not produce racially biased diagnoses,” Vice President Kamala Harris told reporters when the guidance was rolled out.

‘Pick-and-choose process’

A former White House official who worked in the first Trump administration told Nextgov/FCW that they expect the incoming administration to weigh what, if any, parts of the existing implementation guidance to keep and what to wipe away as the new administration evaluates not only the AI executive order, but all of the executive orders signed by the current president.

While parts of the order invoking the Defense Production Act to require companies to hand over information about certain models have come under scrutiny from Republicans, other pieces of the order and guidance to agencies may be less controversial.

Federal agencies only recently released their required plans to comply with the implementation memo for the AI executive order. 

At this point, some agencies may deprioritize continued implementation, said the former White House official. He was also echoed by Divyansh Kaushik, a vice president at Beacon Global Strategies, a national security advisory firm.

The incoming administration may rescind the implementation guidance right away, or they may wait to roll back the memo until they have new guidance to replace it with, the former official said.

That official expects the incoming administration to roll back requirements that are deemed a hindrance to the use of AI and instead speed up the pace of AI adoption across the federal government more broadly and try to compete for AI dominance on the international stage with China.

After the Biden administration released their draft guidance on government use of AI, some AI stakeholders in industry, academia and the public sector alike said that the memo could hamstring the government’s use of AI by miring agencies in red tape.

Sections of the Biden-era implementation guidance focused on high-risk AI applications impacting people’s rights or safety are likely on the chopping block for a close inspection at minimum from the incoming administration, said both the former official and Kaushik, who is expecting a shift away from an anti-bias approach to AI.

A focus on protecting free speech is also to be expected, Kaushik added. 

Trump said he would “ban the use of AI to censor the speech of American citizens” along with reversing Biden’s executive order at an Iowa rally last year.

Anti-bias provisions have drawn the scrutiny of some Republicans already.

A “no woke AI” amendment from Sen. Ted Cruz, R-Texas, passed through a congressional committee in July. It would ban government policies that require or even promote that AI “should be designed in an equitable way that prevents disparate impacts based on a protected class,” among other things.

Artificial intelligence has been proven to demonstrate bias in some cases, even through unintended or unknown biases in the training data used for that AI. 

Trump has also vowed to roll back the current administration’s equity-focused efforts more broadly, calling efforts taken under an equity executive order issued by Biden “a woke takeover of the entire federal government.”

In the event that there’s a vacuum on AI policy, agencies will have discretion on how to move forward, said Robert Shea, who formerly worked in the Office of Management and Budget and is now the CEO of consulting firm GovNavigators.

Such agencies could even continue to implement Biden-era AI governance work if they so choose, unless told to do otherwise. With new guidance from a Trump administration, that discretion would be limited, said Shea.

The first Trump administration did issue some executive orders on the use of AI that are still on the books, so those could also come into play, said Suresh Venkatasubramanian, director of the Center for Technological Responsibility, Reimagination and Redesign at Brown University. He also co-authored the White House Blueprint for an AI Bill of Rights during a stint at the White House’s Office of Science and Technology Policy.

A late 2020 order set out principles for AI in government, as well as public reporting requirements on the use of AI, that the Biden administration built on.

Who leads on AI policy in the Trump administration, as well as who is tapped to head different federal agencies, will likely also shape what agencies do next on AI, said Adam Thierer, resident senior fellow on technology and innovation at the R Street Institute.

Guardrails down

Former Biden officials say that gutting the guidance around the government’s use of AI has its risks.

The worst case scenario is a complete lack of guidance for agencies, said Venkatasubramanian.

“The OMB memo was laying out those guardrails,” he said. “If it is taken down, we will have no protections at all.”

A blank slate for AI governance would also “set the U.S. out of step with where the rest of the world is going,” said Venkatasubramanian, noting that the U.S. could “lose its voice in directing and shaping how AI governance goes.”

Presumably, AI may also be used to further the administration’s own priorities at the border, as Context has reported, and elsewhere.

“Some of the policies that the Trump campaign have proposed and Trump surrogates have proposed that would involve, either directly or indirectly, massive amounts of data collection and analysis by federal agencies,” said Nik Marda, who formerly worked at the Office of Science and Technology Policy in the Biden administration. 

“To not have those government-wide guardrails before the state gets commandeered into potentially egregious violations of civil rights and civil liberties seems really bad,” he said. “These guardrails are also about protecting people from intentional misuses of AI.”

Congressional efforts are also in play as to what rules agencies have to follow when using AI.

The Biden executive order guidance, for example, served a dual function in implementing the AI in Government Act of 2020 and the Advancing American AI Act.

For the upcoming Congress, Republicans have clinched the majority in the Senate. At the time of this article’s publication, it’s unclear who will control the House of Representatives.

Many on Capitol Hill may want to pass some type of AI legislation during the lame duck before the new Congress even comes in, potentially as part of the annual must-pass defense bill. That effort could include government-specific bills, said Mike Hettinger, a former senior congressional staffer who lobbies on behalf of technology companies.

In terms of the government’s use of AI, lawmakers have thrown around a lot of ideas. 

Among them is a bipartisan bill focused on AI procurement from the Senate Homeland Security and Governmental Affairs Committee.

There’s also a bipartisan bill from the House counterpart Committee on Oversight and Accountability that would set up transparency requirements for the use of AI in government decisions and mandate public AI charters for high-risk systems, among other things.

Asked about his legislative agenda on the federal government's use of AI, current Chairman James Comer, R-Ky., pointed to that bill, which he fielded with Rep. Jamie Raskin, D-Md., his Democrat counterpart on the committee.

“The government should focus resources on increasing transparency, oversight and responsible use of federal AI systems while protecting the public’s privacy, civil rights and civil liberties,” Comer said in a statement.

He called Biden's order “overreaching, particularly in its invocation of the Defense Production Act to require domestic developers of frontier AI systems to hand their trade secrets to the government.”

Along with those reporting requirements, work being done at the National Institute of Standards and Technology following the release of the executive order has also come under the fire of some Republicans, as WIRED has reported. 

The future of the agency’s AI Safety Institute — borne out of the executive order broadly but not specifically itemized in it — is also unclear, said Thierer, although there are bills pending to give it statutory backing.

Some, including allies of the former president, have been seeking less harsh regulatory policies from the now-incoming Trump administration, per reporting by the Washington Post, although it is unclear exactly what the approach of the new Trump administration will be.

Regardless of how federal efforts play out, states will likely continue to pass laws on their own, leaving AI policy in a patchwork, said Venkatasubramanian.