OpenAI lobbies blue states for national AI rules

OpenAI has moved from arguing for federal AI rules in Washington to trying to shape them in state capitols run by Democrats. Politico reported on May 20 that the company is lobbying blue-state governments to pass laws that could create momentum for a national framework after federal action stalled. At the same time, President Donald Trump is preparing an executive order on AI and cybersecurity that would set up a voluntary process for companies to share information on advanced models with the government before public release. Chris Lehane, OpenAI’s chief global affairs officer, is at the center of the state push, according to Politico. The report said OpenAI is working state by state after lobbying campaigns in California and New York, a sign that the company is trying to influence the rules emerging outside Congress as lawmakers and regulators debate how to govern powerful models. (politico.com) ### Why is OpenAI working through states instead of waiting for Congress? Washington has been slower to produce a broad federal AI law than many companies and advocacy groups expected. Politico reported that OpenAI’s strategy is to back state measures that align with its preferred national framework rather than wait for Congress to settle the issue on its own. (politico.com) March 2026 already brought one federal marker. The White House published what it called a national AI legislative framework, saying the administration wanted federal leadership on the policy questions raised by AI. That document did not end the debate over oversight, and the administration is now preparing a separate executive order focused on AI and cybersecurity. (politico.com) ### What is Trump preparing to do? Trump is expected to sign an executive order as soon as Thursday, May 21, according to Reuters reporting carried by U.S. News and the Economic Times. The order would establish a voluntary framework for developers to share new models with the government 90 days before public release, according to those reports. (whitehouse.gov) Politico reported that the order would direct a coalition of national security and civilian agencies to increase scrutiny of cutting-edge AI systems. Reuters said White House officials were also trying to assemble AI company chief executives for the signing ceremony. ### What evidence is driving the new pressure for rules? A May 20 Guardian report said a Demos study found chatbots including ChatGPT, Gemini, Grok and Replika made serious mistakes ahead of the Scottish election. (economictimes.indiatimes.com) The report said the systems invented scandals, fabricated candidates and in some cases gave the wrong election date. Britain’s Electoral Commission called for new legal controls, according to the Guardian. (politico.com) In healthcare, a benchmark released May 20 by actAVA.ai tested 30 frontier agents across 75 workflows and found the best-performing systems still failed roughly seven out of ten cases. Press releases carrying the results described the failure rate as 72% across tested U.S. healthcare workflows involving long, multi-step tasks. (vuink.com) ### What does this mean for the companies building the models? OpenAI is now in the position of both building frontier systems and lobbying over the rules that will govern them. Forbes reported earlier this year that OpenAI and Anthropic had increased their lobbying spending as AI policy fights intensified in Washington and the states. (tennessean.com) The immediate next step is in Washington. Reuters reported that Trump could sign the executive order on Thursday, May 21, and Politico said the administration’s order would outline how agencies scrutinize advanced models and cybersecurity risks. (usnews.com) (forbes.com)