OpenAI pushes states for AI rules

Chris Lehane, OpenAI’s chief global affairs officer, is pressing Democratic-led states to pass artificial intelligence rules that mirror the company’s preferred federal framework, according to Politico’s May 20 reporting. The effort comes as Congress remains stalled on a national AI law and as European regulators move ahead with implementation of the European Union’s AI Act. It also lands amid new scrutiny of chatbot reliability after reporting in Britain on false election information generated ahead of Scotland’s parliamentary vote. ### Who is leading OpenAI’s state push? Chris Lehane is the executive driving the campaign. Politico reported on May 20 that Lehane has been pitching what he calls “reverse federalism” — a strategy of shaping national AI policy by securing state-level laws first, rather than waiting for Washington to act. Politico also reported that OpenAI is building on lobbying work in California and New York. (politico.com) The company’s aim, according to the outlet, is to steer emerging state legislation toward a framework it would also like to see adopted nationally. ### Why are states the battleground now? Washington is the immediate reason. (politico.com) Politico described the federal effort on AI safety legislation as stalled, leaving companies and lawmakers without a single national rulebook. In that gap, statehouses have become the place where AI firms are trying to influence the first durable rules. The strategy also reflects how quickly state legislatures have moved. (politico.com) Politico’s newsletter said OpenAI’s plan is to influence bills “one state at a time,” using blue-state capitals as the proving ground for provisions that could become a de facto U.S. standard. ### What is Europe doing at the same time? (politico.com) The European Commission published draft guidance on May 19 to help companies, deployers and regulators decide when an AI system should be classified as “high-risk” under the AI Act. ITPro reported that the Commission simultaneously opened a public consultation on the draft, part of a broader effort to make the law’s obligations more workable in practice. (politico.com) The draft guidance addresses the classification rules under Article 6 of the AI Act and offers examples for how those rules may apply in different sectors and use cases, according to reporting by IAPP and legal analyses that summarized the Commission text. ### Why is chatbot accuracy part of this story? The Guardian reported on May 20 that a Demos study found leading chatbots made serious errors before Scotland’s election. (itpro.com) According to that report, the systems invented candidates, fabricated scandals and gave the wrong voting date, prompting the Electoral Commission to call for new legal controls on AI-generated misinformation. (iapp.org) Separate reporting before the vote also found misleading chatbot answers about constituencies, candidate lists and policy details in the run-up to elections in Scotland and Wales. Those findings have added urgency to arguments from election officials and campaigners that AI systems need clearer safeguards when they are used for civic information. ### What happens next? (vuink.com) The next milestones are split between state capitals and Brussels. In the United States, OpenAI’s lobbying campaign is continuing in legislatures considering AI bills, according to Politico. In Europe, the Commission’s consultation on the May 19 draft guidance will feed into the final interpretation of which systems face the AI Act’s tougher high-risk obligations. (politico.com) (care.org.uk)