Senate starts to fracture over how to govern AI

Cracks are starting to emerge in the Senate’s approach to regulating artificial intelligence — its top tech policy focus of the fall.

On Wednesday, Senate Majority Leader Chuck Schumer convened what he called an “unprecedented” gathering of tech CEOs for an all-day listening session on how the fast-moving technology should be regulated.

But just a day before, Richard Blumenthal (D-Conn.) and Josh Hawley (R-Mo.), two prominent senators on the Judiciary Committee, held a separate hearing to refine their own set of comprehensive AI rules, which they released on Friday. Other senators are busy drafting or introducing bills to address AI-enabled deepfakes, government procurement of automated systems and other piecemeal approaches to the technology.

“Congress operates at the speed of molasses in sub-freezing weather,” Blumenthal told POLITICO on Wednesday. “We need to operate at lightspeed on this issue.”

Blumenthal said that Schumer’s first AI forum was “excellent,” but added that an educational push shouldn’t delay development of real-world legislation aimed at addressing AI.  “You can’t know everything before you do something,” the senator said. “You need to do the legislation and learn at the same time.”

Schumer has dominated the Senate conversation around AI since April, when the majority leader first announced a new effort to “get ahead of” the technology. In June, he put out a broad framework that called on Congress to “join the AI revolution.” Schumer said that roughly two-thirds of senators participated in his day-long session Wednesday.

In contrast to Blumenthal and Hawley, Schumer has taken what he calls a “listen and learn” approach to AI. And he's cautioning his colleagues against getting too specific before Congress has all the facts.

“We don't want to just put together legislation,” Schumer told reporters on Wednesday.

“If you go too fast, you could ruin things,” the majority leader said.

The tension over timing comes as automated systems start to spread into virtually every corner of American life. Maya Wiley, president and CEO of the Leadership Conference on Civil and Human Rights and an attendee at Wednesday’s AI Forum, said AI’s potential to promote biased or discriminatory outcomes means Congress has little time to waste.

“I think there’s no question that these technologies are moving very fast,” Wiley told reporters on Wednesday. “We need the government, and we need legislators, to move quickly. But we need them to move quickly and smartly.”

Fast vs. slow

The emerging Senate split over AI legislation doesn’t break down along partisan lines. In both parties, there are lawmakers urging fast action on a technology that could upend vast swaths of American life – and in both parties, there are lawmakers who believe Congress needs more information before it can move.

“Are we ready to go out and write legislation? Absolutely not,” said Sen. Mike Rounds (R-S.D.), one of Schumer’s three top lieutenants on AI. Schumer said Wednesday that he plans to hold “a bunch more” AI insight forums — an effort that will almost certainly stretch into next year.

But other lawmakers — including Sen. Maria Cantwell (D-Wash.), the powerful chair of the Senate Commerce Committee — say Congress should already have a decent idea of what it would like to accomplish on AI.

“I set up the [National AI Advisory Committee] years ago for this very thing: what should the government role be?” Cantwell told reporters on Wednesday. She argued that Congress has already had “three years” to discuss what AI legislation should “really look like.”

The new AI framework unveiled last week by Hawley and Blumenthal is perhaps the most specific proposal to emerge this Congress. Among other things, it would establish new federal licensing and liability regimes for AI models used in “high risk situations,” as well as new requirements around transparency and kids’ safety.

Among the tech CEOs and civil society advocates at Wednesday’s AI Insight Forum, Blumenthal said there was “very broad consensus — surprisingly strong — in favor of the basic provisions” of his approach. Several key industry voices, including Microsoft president Brad Smith and OpenAI CEO Sam Altman, have already come out in favor of the kind of licensing regimes envisioned by Blumenthal.

Other tech companies, including IBM, oppose the idea.

Some senators say Blumenthal and Hawley’s framework is coming before lawmakers have amassed enough expertise on AI, and could box Congress into a corner.

“My concern as a legislator would be that they get locked into particular provisions as we’re still gathering information, and potentially outflanked by somebody who has had the benefit of more counsel from more sources,” Sen. Todd Young (R-Ind.) told POLITICO on Tuesday.

Young, along with Rounds and Martin Heinrich (D-N.M.), is one of three senators that Schumer has tapped to lead on AI legislation. He added that he “suspects” Blumenthal and Hawley “will be modifying their proposal from time to time.”

More senators, more ideas

Additional AI frameworks are starting to crop up across Capitol Hill. On Wednesday, Sen. Michael Bennet (D-Colo.) wrote to Schumer with elements of his own AI regulatory framework, which contains third-party audits of AI systems, disclosures on AI-generated content and opt-in measures on user data collection for AI training.

Other senators are also moving — including Cantwell, who plans to soon introduce bills that would address the threats posed by deepfakes as well as potential disruptions to jobs and education posed by AI.

On Thursday, Sen. Gary Peters (D-Mich.), chair of the Homeland Security and Governmental Affairs Committee, will hold a hearing on potential legislation to reform Washington’s hidebound procurement process as it relates to AI systems.

Schumer says he’s looking at a committee-by-committee approach to AI regulation as a way to eventually move the Senate toward a more sweeping plan. It’s similar to how Congress crafted what eventually became last cycle’s CHIPS and Science Act (although in that case, lawmakers already had firm ideas about how to reinvigorate high-tech manufacturing and supercharge federal science agencies).

“There's going to be lots of ideas out there,” the majority leader said on Wednesday. “Hawley and Blumenthal have one, many other people have other ideas. It will be our job to put together as comprehensive a plan that we can that A, incorporates many different ideas, and B, can pass.”

Heinrich told POLITICO on Tuesday that going through committees would give the group a sense of "what kind of provisions would garner 60-plus votes, and then we'll have to think about a process for pulling that into some sort of omnibus package sort of situation.”

But other senators, who have seen Congress fail to regulate tech before, are starting to worry that AI could join the list of failed congressional efforts to regulate tech.

“We have not successfully, meaningfully legislated on critical recent emerging technology issues like data privacy,” Sen. Chris Coons (D-Del.), chair of the Senate Judiciary Subcommittee on Intellectual Property, told POLITICO on Tuesday. Coons said he and Sen. Thom Tillis (R-N.C.) will soon release a bill to address AI’s impact on copyright law, apparently with Schumer’s blessing.

“My expectation at this point is that several different committees will take up different pieces of this and try to move forward,” Coons said. “Where it goes after that, I don’t know.”

The elongated timeline of Schumer’s AI forums is causing some to question whether he and his lieutenants actually want to pass comprehensive AI legislation this Congress. That includes Hawley, who on Tuesday told POLITICO he’s “concerned that the leadership of both parties wants to slow down any kind of tech legislation.”

“There’s lots of talk, but the real question is: Are we ever gonna vote?” Hawley said.

Rebecca Kern contributed to this report.