Joe Biden’s Big AI Plan Sounds Scary—but Lacks Bite

Tiny superheroes, fun-size dinosaurs, and overgrown insects squealed at the White House on Monday. The costumed children celebrating Halloween with President Biden weren’t there for the unveiling of a sweeping new executive order on artificial intelligence. Yet as the US government digests its lengthy, new to-do list and Vice President Kamala Harris heads to a UK summit on AI to sell the president’s vision, leaders in Congress and nations around the world may be asking themselves, trick or treat?

While this White House is bullish on the power of the president’s pen, executive orders have limited power domestically—and none overseas. Behind the White House’s rosy PR push about setting a new course for AI lurk the scary but very real monsters of congressional dysfunction and international rivals. Without overcoming both, Biden’s AI vision could struggle to take root as his administration hopes it will.

In Bidenland, where WIRED spent Monday afternoon, everything’s fine though. “As artificial intelligence expands the boundary of human possibility and tests the bounds of human understanding, this landmark executive order is a testament to what we stand for: safety, security, trust, openness, American leadership,” Biden told guests—costumed in scary Washington power suits—gathered in the East Wing of the White House for a signing ceremony. For a moment, another potential government shutdown just 18 days away and China’s determination to leapfrog US technological dominance seemed like just scary stories.

Just as AI is—or soon will be—everywhere, Biden’s executive order touches just about everything the federal government does. It mandates “clear standards to protect rights and safety, improve AI procurement, and strengthen AI deployment” for all government agencies.

Biden even aims to gain a measure of control over private AI projects. He plans to deploy the Defense Production Act—written to allow government control of industries during wartime—to force private US technology firms to report sensitive details of their most secretive AI development projects to the federal government.

“This executive order will use the same authority to make companies prove—prove—that their most powerful systems are safe before allowing them to be used,” Biden said.

Vice President Harris was at his side for the announcement but is taking his AI vision on the road for the rest of the week. She’ll be taking her own agenda to the UK’s Summit on AI Safety, hosted by Prime Minister Rishi Sunak and primarily focused on far-off AI risks.

“We intend that the actions we are taking domestically will serve as a model for international action, understanding that AI developed in one nation can impact the lives and livelihoods of billions of people around the world,” Harris said Monday. “Fundamentally, it is our belief that technology with global impact requires global action.”

Ahead of Biden’s address, WIRED sat down with Arati Prabhakar, director of the White House Office of Science and Technology Policy. She was quick to bat away a question about who has the president’s ear on generative AI. “Let's start at the beginning,” she offered. “This is the biggest action anyone's ever taken anywhere in the world on AI, and the president's driven it from the beginning.”

Prabhakar also didn’t much want to talk about how a law designed for wartime could be casually used to enact new tech regulation. “The Defense Production Act has been used in times of crisis when we face significant national security issues,” Prabhakar said, noting its use to speed up delivery of vaccines and PPE during the pandemic.

But is the US—WIRED asked in countless ways—in a national crisis caused by AI on the level of a war or global pandemic?

“The national security issues are not the whole story but one part that does require paying attention to,” Prabhakar said. “That's the reason for using the Defense Production Act for the very specific purpose of getting notification and disclosure related to powerful model development beyond where we are today.” It’s a nice theory that could end up being put to the test in court, if companies challenge this new use case for a law generally tapped only in emergencies.

Prabhakar steered the conversation back toward more everyday AI but ended up highlighting other limitations of Biden’s latest executive action. “The vast majority of actions in this executive order are about how we use the AI technologies that are already out in the world responsibly,” she said. “How do we make sure they don't violate Americans' privacy? How do we make sure that we don't embed bias that changes, you know, where people can live and whether they get a loan or whether they go to jail or not?”

Great questions—but changing the rules for privacy or mandating, say, the universal watermarking of AI-altered images for private companies requires legislation from Congress. Indeed, even as his White House is going all out to sell the expansive new executive order as a historic first, Biden admits that the order falls short of what the changes wrought by this ever-evolving technology demand. “This executive order represents bold action, but we still need Congress to act,” Biden said.

Senate majority leader Chuck Schumer know’s that too. His job of late has been to radiate optimism at the US Capitol despite being unable to pass even broadly bipartisan tech regulation aimed at protecting children’s online privacy in recent years. Given the history of congressional inaction on protecting Americans’ data, WIRED asked him at a press briefing after the signing ceremony whether having the White House move, with a flip of a presidential wrist, the “biggest action anyone's ever taken anywhere in the world on AI” takes pressure off a complacent, if not incompetent, Congress.

“No,” Schumer replied. “It importunes us to act, because lots of the things that were done here can only be done with federal contracts. “For most of this, legislation is required to cement it, to expand it and to make sure everybody obeys it.”

Last week, Biden met with the new, Republican, Speaker of the House Mike Johnson for the first time. Schumer has yet to. Still, at the White House Schumer promised a new AI bill soon. He and a bipartisan group of four senators—the ones who’ve been hosting all-Senate AI briefings and forums—are meeting with Biden at the White House this week. Schumer cranked up the optimism. "It won't be days or weeks, but it won't be years,” he said, urging his audience, in case of any doubt, to expect it in “months.”

That’s months away from introduction, not passage. Vice President Harris is unlikely to lead with that as she tries to project strong leadership on AI on the international stage in London this week.