OpenAI seeks transparency rules

OpenAI is trying to shape the next phase of U.S. AI regulation through state legislatures rather than waiting for Congress. Politico reported on May 20 that the company is pushing states to adopt laws built around transparency and reporting requirements for advanced AI developers, while avoiding steep penalties and new liability for catastrophic harms. Chris Lehane, OpenAI’s chief global affairs officer, told Politico the strategy is meant to create a de facto national standard by aligning rules across large states. (politico.com) The campaign comes as the Trump administration weighs new federal action on frontier AI models. CNN reported that the White House could issue an executive order as soon as Thursday, May 21, that would push for a voluntary government review of advanced models before public release. Reuters, cited by U.S. News, separately reported that Trump was expected to sign an order on AI and cybersecurity as soon as Thursday. (politico.com) ### Which rules is OpenAI backing in the states? Politico reported that OpenAI is advocating laws centered on transparency and reporting obligations for top AI developers. Those proposals, according to the report, are modeled on measures passed in California and New York late last year after what Politico described as substantial input from OpenAI lobbyists. Chris Lehane told Politico that OpenAI wants “a bunch of the big states to come together and mirror each other” to create a national standard in practice. (kesq.com) Politico said the company’s preferred approach would give it a more stable legal framework while exposing it to relatively little new liability for catastrophic harms. ### Why is the company working state by state instead of through Congress? Congress has not produced a consensus federal AI framework, and Politico said that deadlock is driving OpenAI’s state strategy. (politico.com) Lehane described the approach as “reverse federalism,” with the company seeking to shape a critical mass of state laws that could function as a national baseline. Illinois is the next target in that effort, according to Politico. (politico.com) The outlet reported that Lehane said OpenAI had already found success in California and New York and was now looking to Illinois as lawmakers there move legislation that mirrors those states’ rules. Politico also reported that OpenAI has been working not only legislatures but governors’ offices, including those of Gavin Newsom in California and Kathy Hochul in New York. ### What changed in Washington? Anthropic’s Mythos release has become part of the backdrop for the White House debate over AI oversight. Politico reported on May 7 that Anthropic’s model was being made available only to a small group of threat researchers and tech companies because of its advanced hacking capabilities, and that OpenAI had previewed its own GPT-5.5-Cyber model for White House and Commerce Department officials. (politico.com) CNN reported that the administration’s thinking hardened after concerns about frontier models’ cyber capabilities. The proposed executive action, according to CNN, would include a voluntary arrangement under which AI companies share advanced models with the government for a period before launch. Reuters reported that the White House was also trying to gather AI company chief executives for a signing ceremony. (politico.com) ### Why does Singapore appear in the same story? Singapore announced a separate agreement with OpenAI on May 20 as the company expanded overseas. CNBC reported that OpenAI signed its first memorandum of understanding with Singapore and committed more than 300 million Singapore dollars, or $234 million, to strengthen the country’s AI ecosystem. The deal includes OpenAI’s first overseas AI lab, in Singapore, and the lab is expected to employ more than 200 people over the next few years, CNBC reported. (kesq.com) The partnership is focused on areas including education, public services, finance, healthcare and digital infrastructure, according to the joint statement cited by CNBC. ### What happens next? May 21 is the next immediate date to watch in Washington because the White House could move on an executive order covering advanced AI model review. (cnbc.com) Illinois is the next named state in OpenAI’s lobbying campaign, according to Politico, while Singapore’s new applied AI lab will begin building out staffing under the memorandum announced this week. (kesq.com)