WASHINGTON — Sen. Maria Cantwell on Wednesday proposed a federal framework requiring safety standards and pre-release reviews for the most advanced artificial intelligence systems, renewing pressure on Congress to act as the technology develops faster than lawmakers have agreed on rules to govern it. The Washington Democrat’s principles would put enforceable standards at the center of U.S. AI policy. The National Institute of Standards and Technology would lead development of those standards with other federal agencies, researchers and leading AI companies. Advanced systems would face government testing, independent audits or both before release, with continuing scrutiny as the technology changes. The proposal does not spell out how independent auditors would be selected, what penalties would apply when companies violate standards, or how quickly developers must report safety incidents. Cantwell described the principles as a starting point for discussions, rather than a finished regulatory plan. “These are very comprehensive, in the way that the system could work,” Cantwell said in an interview. “But now we have got to get people to dialogue about that.” Her plan calls for a national network linking advanced AI developers, universities, independent researchers and the federal government. Standards would address risks including cyberattacks and the possibility that people could lose control of powerful systems. Developers would have to provide access to testers and auditors, and report safety incidents. The principles also promote corporate transparency, public-private research partnerships, protections for children online and international cooperation. They call on AI companies to support education, apprenticeships and worker training amid concerns about job displacement. The framework does not address environmental harms, though Cantwell said they should be included. It also proposes a secure communication channel with China for major AI incidents that could threaten international security. The framework comes amid competing proposals from lawmakers who broadly support federal oversight but disagree on its form. Sen. Bernie Sanders of Vermont and Rep. Greg Casar of Texas have proposed banning superintelligent AI and temporarily pausing advanced development under a new federal agency. Rep. Pramila Jayapal of Washington has proposed a federal charter system for AI companies modeled on bank regulation. A separate bipartisan effort by Senate Majority Leader John Thune and Sen. Amy Klobuchar has not yet produced bill text. Congress has not advanced comprehensive AI legislation this year. President Donald Trump has opposed new regulation and dismissed warnings about catastrophic AI risks as a “hoax.” Cantwell argued that calls from some technology companies for federal standards could help lawmakers find common ground, after an earlier bipartisan AI bill failed to reach a Senate vote. The proposal largely builds on ideas in a 2024 bill Cantwell developed with Republican Sen. Todd Young. That measure proposed voluntary standards and assessments, but stalled in the Senate. Cantwell’s new principles instead call for mandatory standards for the highest-risk systems. Some AI safety experts have urged Congress to move beyond a framework and establish specific, rigorous requirements. They differ on whether oversight should sit in a new federal agency, whether development should slow until safeguards are in place, and whether government or private auditors can best ensure independent scrutiny. Experts have also raised privacy and environmental concerns. “Unless we act soon, we are actively losing control — and by we, I mean the public and policymakers, everyone who’s not inside of the company,” said Stephen Casper, a computer scientist and assistant professor at Harvard Kennedy School. The timing is a challenge. Congress is not scheduled to reconvene until after the November midterm election, leaving lawmakers few days together in session when they return in mid-November. Cantwell said short-term protections addressing the largest risks might be possible before the end of the year, while broader legislation could advance next year. She did not provide a timeline for creating the proposed standards. Cantwell could gain greater influence over the issue if Democrats win control of the Senate in November. She is positioned to take over the Senate Commerce, Science and Transportation Committee in January from Republican Sen. Ted Cruz, who has not held markups on bipartisan AI safety legislation this year. Senate Democratic leader Chuck Schumer endorsed Cantwell’s principles, saying they could lay groundwork for legislation that protects the public while supporting AI development. Microsoft President Brad Smith also backed the proposal as a contribution to the national debate.