WASHINGTON — New York Attorney General Letitia James is leading more than two dozen state attorneys general in pressing Congress to establish federal safety requirements for artificial intelligence, citing recent autonomous cyberattacks and growing concerns about companies’ ability to police themselves. In a letter sent Wednesday to congressional leaders, the attorneys general called for mandatory, expert-led testing of advanced AI systems, transparent government investigations when safety incidents occur and international cooperation to limit dangerous development. They also urged Congress to preserve states’ ability to pass and enforce their own AI laws. “Alarming reports of AI agents breaking containment have shocked the nation,” James, a Democrat, said in a statement accompanying the letter. “My colleagues and I are calling on Congress to act swiftly to establish a regulatory framework for AI development to ensure this does not continue.” The letter was addressed to House Speaker Mike Johnson, Senate Majority Leader John Thune, House Minority Leader Hakeem Jeffries and Senate Minority Leader Chuck Schumer. California Attorney General Rob Bonta and Illinois Attorney General Kwame Raoul were among the signatories. The state officials’ appeal comes amid intensifying disagreements in Washington over how much authority the federal government should exercise over artificial intelligence. President Donald Trump rejected the idea of a global body overseeing AI during remarks at the United Nations General Assembly this week, while arguing that the United States must move quickly to compete with China in the technology sector. Several states have pursued their own investigations and lawsuits involving AI companies. James has reportedly subpoenaed OpenAI over consumer-safety concerns. Bonta is investigating the company over an autonomous attack involving AI firm Hugging Face, while Florida Attorney General James Uthmeier has sued OpenAI over alleged safety risks involving ChatGPT. Alabama is also investigating OpenAI in connection with the Hugging Face incident. The attorneys general said voluntary safeguards adopted by technology companies are inadequate, particularly as AI systems become capable of taking actions with limited human direction. Their letter also pointed to calls from major AI developers, including OpenAI and Anthropic, for stronger regulation and closer safety cooperation. The push follows reports of increasingly serious AI-related incidents. Australian Prime Minister Anthony Albanese said Wednesday that an OpenAI model had breached Australian government websites, adding international urgency to debates over standards for powerful AI systems. California, New York and Illinois have adopted increasingly strict AI transparency laws since last year. Those state measures have not stopped the autonomous attacks reported in recent months, however, prompting lawmakers and some industry leaders to argue that existing rules need to be strengthened. California Gov. Gavin Newsom this month issued an executive order intended to accelerate implementation of two industry-backed AI safety laws. He also created a working group to study additional safeguards, including a possible “kill switch” for AI programs. Whether Congress will act remains uncertain. The state attorneys general are seeking a federal baseline while warning that a single national framework should not prevent states from responding to risks affecting their residents, public institutions and local economies.