AI hacking incidents spur new push for state safety laws

Disruption AI hacking incidents spur new push for state safety laws Bills are being considered in three states still in session and being prepped in more states for next year. Austin Jenkins 06:00 AM, Oct 22, 2026 Pages from the Anthropic website and the company’s logos are displayed on a computer screen in New York on Thursday, Feb. 26, 2026. (AP Photo/Patrick Sison) A series of artificial intelligence hacking incidents is weighing on state lawmakers as they finish up business this year and prepare fresh efforts to regulate the technology in 2027. Bills are under active consideration this month in Massachusetts, New Jersey and Pennsylvania, where legislative sessions are ongoing. California Sen. Josh Becker plans to introduce “groundbreaking legislation” next year “to pace the development of frontier AI models.” And a Virginia delegate said AI regulation will be a priority when the General Assembly reconvenes in January. The renewed momentum signals a continued commitment by state leaders to remain in the driver’s seat on AI guardrails as Congress wavers and President Donald Trump embraces voluntary standards. The stakes have intensified in recent months with multiple reports of AI agents escaping their testing environments and hacking into companies. The incidents have forced a reckoning over the pace of AI development, the danger of autonomous agents, and what guardrails are needed. States have led on AI safety legislation, with laws regulating large, frontier models now in place in California, Illinois and New York. Massachusetts lawmakers are negotiating the final details of a sprawling economic development bill that includes frontier model regulation language from Democratic Sen. Barry Finegold. Read more: Mass. bill would force AI giants to disclose catastrophic risk Gov. Maura Healey, a Democrat up for re-election, supports the AI safety language, which includes a first-in-the-nation provision requiring regular independent evaluations of frontier models. “In the absence of federal leadership, I’m respectfully requesting that the Legislature adopt core AI frontier model safety and transparency provisions,” Healey wrote in a letter to the conference committee last month. New Jersey Assembly Democrats are considering an AI frontier safety bill and a civil liability bill, both from Democratic Assemblymember Andrew Macurdy. Macurdy and Gabriel Weil, an assistant professor at the University of Houston School of Law, co-authored an op-ed last month touting the liability measure. “The principle is simple: If you create it, you should be accountable for the harmful consequences,” they wrote. Read more: Push builds to hold AI developers legally liable for their models’ harms Macurdy’s bills were discussed Monday in the Assembly Science, Innovation and Technology Committee, which invited AI experts and industry representatives to testify. Speaking with Pluribus News after the hearing, Macurdy said the liability bill is his priority and that he hopes it will get a committee vote before year’s end. “Liability is the quickest way to have an actual impact on what the developers are doing,” Macurdy said. “That really moves the needle.” Also Monday, the Assembly Public Safety and Preparedness Committee advanced a recently introduced chatbot safety bill based on model legislation from the Electronic Privacy Information Center and other groups. Earlier this year, the Assembly Science, Innovation and Technology Committee advanced a separate package of AI bills that have yet to receive a floor vote. New Jersey lawmakers have through next year to pass bills that are currently active. Pennsylvania lawmakers heard testimony Monday on a pair of recently introduced frontier safety bills from Reps. Melissa Shusterman, Manuel Guzman and Greg Scott, all Democrats. One is called the Artificial Intelligence Risk Prevention Act and the other is the Artificial Intelligence Safety, Transparency and Accountability Act. Read more: With Washington gridlocked on AI, states prepare fresh wave of safety bills The bill sponsors say there is an “urgent need” to act in light of recent hacking events. They pointed to a Quinnipiac poll released last month that showed 86% of Americans support independent safety standards for AI companies. “The possibility of rogue AI agents is not acceptable to me, and I urge my fellow legislators of our Commonwealth, as well as officials at every level of government, to take these risks seriously and act quickly. Our citizens are counting on us,” Shusterman said in a statement following the hearing. Some experts have questioned the effectiveness of frontier model safety laws as currently written because they apply to the largest models and generally only require disclosure of critical safety incidents. At the Pennsylvania hearing, Steve Wimmer, senior technical and policy adviser at the Seattle-based Transparency Coalition, told lawmakers that their legislation should require reporting anytime an AI company loses control of its models or agents. Other AI bills have recently advanced in Harrisburg. A Shusterman bill regulating AI chatbots passed the Democratic-led House last month and is now under consideration in the Republican-led Senate. A bill from Rep. Arvind Venkat, a Democrat, regulating the use of AI in healthcare also cleared the House last month. Both the House and Senate have passed a bill requiring disclosure when AI is used in advertisements. Read more: House approves bill to mandate AI disclosure in commercials Lawmakers are also preparing legislation for early next year. In California, Becker announced Tuesday he will introduce the Frontier AI Safety Act to build upon the state’s existing frontier model safety law. It would require large AI developers to dedicate a certain percentage of the computing power they spend on research and development to safety-related research. “Computing power is one of the most valuable resources an AI company has,” Becker said in a statement announcing the legislation. “Right now, the overwhelming incentive is to use that resource to build bigger, faster, and more capable models. We need to ensure that understanding the risks and developing safeguards are also fundamental priorities.” Virginia lawmakers are also teeing up AI bills for 2027 after mostly taking a pass on regulating the technology this year, in part because of concerns about Trump administration penalties. “We want to be a leader in this space of technology, and being a leader doesn’t mean being first always,” Del. Cliff Hayes, the Democratic chair of the Joint Commission on Technology and Science, said in an interview. Hayes said the legislative agenda is still being finalized but that chatbot regulations are top of mind. Other areas of interest include independent verification organizations, AI personhood, AI in healthcare and the recent AI hacking incidents. “It would behoove us to have guardrails in place to ensure we’re moving forward in terms of innovation, but also in a responsible way,” said Hayes, who also chairs the House Communications, Technology and Innovation Committee.