The House Ways and Means Committee voted 24-21 Wednesday to reject a sweeping moratorium on new large-scale AI model deployments, killing for now a bill that had passed the Senate two weeks earlier and dealing a sharp blow to a coalition of progressive lawmakers, labor unions and consumer-safety groups who had spent the spring pressing Congress to pause the industry’s headlong expansion.

The legislation, co-authored by Sen. Bernie Sanders, I-Vt., and Rep. Alexandria Ocasio-Cortez, D-N.Y., would have imposed an 18-month freeze on the public release of frontier AI systems above a defined compute threshold, paired with mandatory disclosures on training data, energy draw and labor displacement. It cleared the Senate 52-48 on April 7. In committee Wednesday, four Democrats joined every Republican in voting it down.

Within an hour of the gavel, Sanders called the result “a victory bought and paid for by the largest corporations in human history” and pledged to reintroduce a revised version before the August recess. “We are not going away,” he told reporters in a Capitol hallway. “If the United States Congress will not regulate the most powerful technology ever invented, the states will, and the voters will remember in November.”

Ocasio-Cortez, who had spent much of the past two weeks shuttling between freshman House members and skeptical committee staff, was more pointed. In a video statement posted from her office shortly after the vote, she singled out the lobbying campaign that has descended on Washington since the Senate vote, calling it “the most expensive influence operation I have ever witnessed in this building.” She said her office had logged more than 300 meeting requests from hyperscaler executives, trade associations and law firms representing AI developers in the past 14 days alone.

“They spent more to kill this bill than most school districts spend in a year,” Ocasio-Cortez said. “And they spent it because they knew it would work. That should tell every American something about who actually runs this town.”

The committee’s chairman, Rep. Daniel Holcombe, R-Tex., framed the vote as a defense of American competitiveness, arguing the moratorium would “hand the next decade of artificial intelligence to Beijing.” In remarks on the House floor following the vote, Holcombe said the committee would instead take up “a series of targeted measures” addressing energy consumption at data centers, transparency in training data and a federal pre-emption framework that would block what he called “a patchwork of unworkable state rules.”

The four Democrats who broke with their party — three from districts hosting major data center construction and one from a Silicon Valley-adjacent suburb — issued a joint statement defending the vote as “a recognition that an 18-month freeze is the wrong tool for a real problem.” They said they remained open to negotiating a narrower bill focused on disclosure, child-safety guardrails and grid-impact reporting.

Industry groups, which had warned for weeks that the moratorium would devastate U.S. positioning against Chinese model developers, responded with carefully calibrated relief. The American Frontier Technology Council, a trade association whose members include the country’s four largest model developers, issued a statement praising the committee for “rejecting a blunt-force approach” while acknowledging “legitimate public concerns about energy, labor and safety that the industry must help address.”

Privately, executives were less measured. “This was existential,” said Marcus Whelan, a senior policy adviser at one of the hyperscalers, speaking on condition of anonymity because he was not authorized to discuss internal lobbying strategy. “We weren’t going to let a Senate vote written by Bernie Sanders set the regulatory floor for the next decade. The House did its job.”

The bill’s defeat arrives at a moment of mounting public unease about the pace of AI deployment. A Quinnipiac survey released Tuesday found 58 percent of registered voters supported “a temporary pause” on the largest AI models, with majorities of both Democrats and independents in favor and Republicans split. Labor unions, including the AFL-CIO and the Writers Guild of America East, had endorsed the legislation, citing accelerating job displacement in administrative, creative and customer-service roles.

Sanders and Ocasio-Cortez signaled Wednesday they would now pivot to a two-track strategy: a slimmed-down federal bill aimed at the disclosure provisions that drew bipartisan interest in the Senate, and an aggressive push for state-level moratoriums. Bills modeled on the federal text are already advancing in the New York and California legislatures, and Sanders said sponsors in five additional states had reached out to his office in the past week.

Energy concerns may yet force the issue regardless of the committee outcome. Grid operators in Virginia, Texas and Arizona have warned in recent filings that data center demand is outpacing generation capacity, and the Federal Energy Regulatory Commission is weighing new interconnection rules that could effectively delay some of the deployments the moratorium sought to pause. Asked whether that regulatory pressure could substitute for legislation, Sanders shook his head. “FERC can slow them down. It cannot make them tell the truth about what they’re building or what it will do to working people,” he said.

The White House, which had publicly declined to take a position on the bill, issued a brief statement Wednesday evening saying President Trump “looks forward to working with Congress on a balanced framework that protects American leadership in artificial intelligence.” Administration officials said they expected to outline their own legislative priorities on the issue within the next several weeks.