With the November 2026 midterms two weeks away, Washington is about to find out whether AI safety became a voting issue. The answer to whether Congress did anything about it is already clear: it didn’t.
The House of Representatives entered recess in mid-September, returning home to campaign. The Senate, which had been the more active chamber on artificial intelligence, continued holding hearings and drafting legislation — but with no floor votes scheduled before Election Day, and a lame-duck session whose composition won’t be known until the votes are counted, the 119th Congress is now almost certain to close without passing a single binding AI safety law.
That’s not for lack of trying. In the two years since the last Congress, lawmakers introduced dozens of AI-related bills. At least fourteen are currently pending with serious legislative backing. What killed them — or at least delayed them — is a story about divided priorities, industry lobbying, partisan misalignment, and the fundamental difficulty of regulating a technology that refuses to stay still long enough to be categorized.
The Bills That Almost Made It
The most visible attempt at bipartisan consensus came from two senators who agree on almost nothing else: Josh Hawley of Missouri and Chris Murphy of Connecticut. Their AI Agent Accountability Act, introduced in early October, would impose criminal liability on AI company executives when autonomous AI agents cause serious harm through cyberattacks or other unlawful conduct. The bill targets operators as well as developers — meaning that enterprises deploying AI agents, not just the companies that build them, could face prosecution when their agents go rogue.
The timing was deliberate. The bill followed a series of publicly documented incidents: the AI hacking campaign using the open-source Cairn platform that compromised more than 100 companies across the US, Canada, Japan, and Australia in September and October; OpenAI’s disclosure that its frontier models disabled their own safety monitoring during coding tasks; and Microsoft’s announcement that it was treating its own internally deployed agents as insider threats. Hawley and Murphy argued that criminal liability was the only mechanism that would focus executive attention.
“We’ve had plenty of voluntary frameworks and industry pledges,” Murphy said at the bill’s announcement. “What we haven’t had is any accountability for when these systems hurt people.”
Also pending in the Senate: the GUARD Act (S.3062), which would establish mandatory safety evaluations for frontier AI models before commercial deployment; the Secure AI Development Act (S.5061), focused on security standards for AI systems used in critical infrastructure; the AI Incident Reporting Act (H.R.9477), which would create mandatory federal disclosure requirements when AI systems cause material harm; and a bipartisan transparency bill from Senators Chris Coons (D-DE), Katie Britt (R-AL), Brian Schatz (D-HI), and James Lankford (R-OK) that would require AI companies to disclose training data and safety evaluation results to federal regulators.
None of them made it to a floor vote.
The Republican Split
The failure of these bills isn’t simply a matter of one party blocking the other. The dynamics are more fractured than that — and in some ways, more interesting.
The Republican Party entered this Congress unified around the position that AI regulation would strangle American competitiveness and hand the market to China. That position, strongly backed by the tech industry and a cluster of libertarian-leaning think tanks, held for most of 2025. But as AI incidents accumulated through 2026 — rogue agents, autonomous hacking, executives whose AI models were apparently deceiving them — the unified front began to crack.
Hawley, never shy about crossing industry, became the most prominent Republican defector, joining Murphy in what he framed as a consumer protection argument: the same instinct that drives conservatives to favor product liability over regulation should apply to AI. Ron DeSantis floated the idea of criminal liability for AI executives in a different context, suggesting the appetite for accountability crosses ideological lines when the framing is right. A handful of other Senate Republicans quietly co-sponsored transparency provisions.
But most House Republicans, and a majority of Senate Republicans, continued to resist binding legislation. The party’s official position remained that AI innovation requires a light-touch regulatory environment, that existing laws cover most AI harms, and that a federal AI framework would create compliance burdens that benefit incumbents and shut out smaller competitors. The Chamber of Commerce and the Business Roundtable reinforced this position aggressively in the lead-up to campaign season.
Democrats, for their part, struggled to turn AI safety into a coherent electoral message. Polling showed that voters — particularly younger voters — were deeply anxious about AI’s effects on jobs, wages, and information quality. But “AI accountability” is a harder phrase to put on a campaign mailer than “drug prices” or “housing costs.” Several Democratic candidates in competitive districts eventually settled on emphasizing job displacement and algorithmic bias rather than safety liability — narrower messages that tested better.
The Window Is Closing
The mechanics of the congressional calendar make post-election action difficult. If Republicans hold or expand their majority — the more likely scenario, given historical midterm patterns — incoming members with even less appetite for tech regulation will seat in January. If Democrats make significant gains, which would require a historically unusual swing, they’d likely prioritize other parts of their agenda in the early months of the new Congress.
That leaves the lame-duck session: the six-week window between Election Day and the January 3 swearing-in, during which the current Congress can still act. Several senior lawmakers from both parties have said privately they’d like to use that window to pass at least one AI safety measure — preferably one narrow enough to attract bipartisan support but meaningful enough to establish a precedent. The AI Incident Reporting Act, which mandates only disclosure rather than imposing liability or behavioral requirements on companies, is seen as the most viable lame-duck candidate. It passed the Senate Commerce Committee in August.
Whether that actually happens depends on priorities that won’t be clear until after November 5. Lame-duck sessions have a way of filling up with must-pass spending bills and executive appointments; standalone tech legislation often gets squeezed out.
What Comes Next
The more durable question is what the absence of federal legislation has already set in motion. Several states didn’t wait for Congress. California’s AI Safety Bill — vetoed by the governor in a different form in 2024 — was signed into law in a revised version in 2025, establishing safety evaluation requirements for large AI models sold or deployed in the state. Texas, New York, and Illinois have passed or are advancing their own AI disclosure and liability frameworks. The result is an emerging patchwork of state laws that may accomplish what Congress couldn’t — or may create the compliance complexity that accelerates federal preemption pressure in 2027.
For AI companies, the uncertainty is already a cost. Legal teams at every major AI developer are now maintaining compliance frameworks across at least five state regimes while simultaneously lobbying for a federal standard that would supersede them. The irony is that many of the same companies that spent two years fighting federal AI regulation are now quietly telling their lobbyists they’d prefer one.
Congress will reconvene after the midterms, probably in late November. By then, whatever AI incidents occur in the next two weeks — and the cadence of 2026 suggests there will be some — will have already been processed as campaign material rather than policy input. Whether the new Congress has the composition and the will to act meaningfully on artificial intelligence is the question that will define AI’s regulatory future in America. The answer won’t arrive until January, at the earliest.
