WASHINGTON — Senate Majority Leader John Thune is trying to broker bipartisan legislation aimed at placing safety requirements on the most powerful artificial intelligence systems, but disagreements within both parties are putting the effort in jeopardy. Thune, a South Dakota Republican and former chair of the Senate Commerce Committee, said Monday that passing an AI bill by the end of the year remains uncertain. He said lawmakers face competing priorities and that reaching an agreement soon would be “challenging.” “There are going to continue to be a lot of conversations about potential paths forward,” Thune told reporters. “But getting anything done in the near term is going to be challenging given the other stuff we’re dealing with.” The legislation under discussion, developed by Thune and Sen. Amy Klobuchar of Minnesota, would establish safety and security standards for the AI models considered most dangerous. Thune has promoted federal action on the technology for years, including an earlier version of the proposal introduced in 2023. But the negotiations have become more difficult as lawmakers debate whether federal rules should override state AI safety laws. Many Republicans argue that overly prescriptive regulation could slow American innovation and allow China to gain an advantage. Thune said Congress should take a “light touch,” while still addressing the most serious threats. “There is a way in which Congress can put guardrails around those types of more consequential threats, without in any way harming our ability to stay ahead in the AI race,” Thune said. That approach is drawing criticism from Democrats who say the industry’s rapid expansion demands stronger oversight. Sen. Maria Cantwell of Washington, the top Democrat on the Commerce Committee, said Republicans have repeatedly urged restraint even after experts warned that AI systems could become uncontrollable. When told Thune had again endorsed a light-touch approach, Cantwell responded: “Okay, that’s bad.” Klobuchar said Monday that negotiations are continuing and that lawmakers broadly agree on the need for legislation. She said she wants stronger safeguards and does not expect the committee’s chair and ranking member to agree on every provision. “Time is of the essence,” Klobuchar said in a statement. Senate Republicans were expected to discuss AI policy at a closed-door lunch Tuesday. The meeting comes after an industry researcher warned that advanced AI could threaten humanity and several company leaders called for a coordinated slowdown in the development of increasingly powerful models. Sen. John Barrasso of Wyoming, the Senate’s No. 2 Republican, said lawmakers have been discussing AI for years but would now address the issue “more aggressively.” The debate is also unsettled in the House, where Speaker Mike Johnson and other Republican leaders have warned against moving too quickly on regulation. Some small-government conservatives, however, argue that Congress should establish basic protections rather than leave the industry without federal standards. Sen. Ted Cruz of Texas, the Republican chair of the Commerce Committee, has acknowledged that warnings from AI laboratories are “scary as hell.” Yet the committee has repeatedly delayed votes on AI legislation, and Cruz said Monday that lawmakers were still negotiating and had not reached an agreement on the Thune-Klobuchar proposal. Democrats are divided as well. Sen. Bernie Sanders, an independent from Vermont who caucuses with Democrats, has called for an immediate pause in the development of AI systems capable of superintelligence. Sen. Brian Schatz of Hawaii has advocated a narrower first step: requiring transparency, outside safety certification and oversight before advanced models are released. “What we need to do is take down the risk of these models either jumping out of the sandbox or being released into the wild without any third party certifying to their safety,” Schatz said. The timing is tight. The House is expected to leave Washington this week until after the November elections, and the Senate could adjourn as soon as next week. Neither chamber is expected to consider broad AI safety legislation before the recess. A separate measure addressing the energy demands of data centers could advance before then, but only if all 100 senators agree to expedite it. The issue has growing significance for communities confronting new data center construction, including questions about electricity use, grid capacity and environmental impacts. For now, Thune faces pressure from both sides: Republicans who fear regulation will undermine the technology industry and Democrats who argue that voluntary safeguards have not kept pace with the risks. Whether the Senate can produce a compromise may determine how much responsibility remains with states — and how much power Congress gives federal regulators over the companies building the next generation of AI.