Democrats push AI safety framework as Congress stalls on regulation

Senator Maria Cantwell rolled out a framework for artificial intelligence safety this fall as Congress remained stuck on the larger question of how—or whether—to regulate the technology at all. The stalemate in Washington came as the election cycle pressed forward, leaving the field open for individual states to move ahead with their own rules, a prospect that prompted fellow Democrat Patty Murray to convene conversations with Washington state officials about what federal guardrails might look like.

Cantwell's proposal arrived into a legislative vacuum. Congress had spent months circling around AI oversight without landing on a coherent approach, and the political calendar offered little incentive to break the impasse before voters went to the polls. The framework she unveiled was meant to address safety concerns that had animated tech policy debates for years: how to ensure that increasingly powerful AI systems don't cause harm, how to audit their behavior, how to keep bad actors from weaponizing them. But a proposal from a single senator, however detailed, could not move the needle on its own.

Murray took a different tack. She traveled to Olympia to sit down with state lawmakers and administrators who were grappling with AI regulation on their own. The message she carried was pointed: states needed federal action, but not the kind that would lock them out of their own policymaking. She was pushing back against a common industry argument—that a single national standard would be cleaner and cheaper than a patchwork of state rules. Murray's position suggested that Washington state, and others like it, should have room to set their own floors and ceilings even as the federal government established baseline protections.

In the House, Representatives Carter and Thompson reintroduced legislation that would task the Cybersecurity and Infrastructure Security Agency with standing up a dedicated force to tackle AI-related threats. The bill had been introduced before; its reintroduction signaled that at least some Democrats were committed to the idea even if the broader legislative machinery had stalled. The CISA task force approach represented a narrower, more focused intervention than Cantwell's broader safety framework—it zeroed in on the security dimension of AI rather than trying to solve the entire regulatory puzzle at once.

What tied these efforts together was a shared diagnosis: Congress was not moving fast enough, and the vacuum was being filled by states, by companies themselves, and by the absence of any coherent national standard. Democrats were trying to thread a needle—pushing for federal action without suggesting that federal action should preempt everything states wanted to do. It was a delicate position, one that reflected genuine uncertainty about what the right regulatory model even looked like. The technology was moving faster than lawmakers could think through the implications. Election season made it harder still to build the kind of bipartisan consensus that major legislation typically required.

By October, with the election weeks away, it was clear that any major AI regulation would have to wait for the next Congress. What remained was the groundwork: frameworks floated, task forces proposed, state officials convened. Whether any of it would translate into law depended on what happened at the ballot box and whether the incoming Congress would treat AI oversight as urgent enough to move quickly.