Trump eyes AI executive order

President Donald Trump is expected to sign an executive order on artificial intelligence and cybersecurity as soon as Thursday, according to Reuters and CNN, after weeks of White House debate over how to monitor powerful new AI systems. The draft order under discussion would push companies to submit advanced models for a voluntary government review before public release, CNN reported. The move comes after Anthropic unveiled Mythos, a model the company said could identify and exploit software vulnerabilities at unusual speed. At the same time, OpenAI has been pressing lawmakers for transparency and reporting rules that would impose fewer liabilities than some AI safety advocates want. ### What would Trump’s order do before new AI models are released? CNN reported on May 20 that the White House was preparing an order built around a voluntary review process for advanced models before launch. Reuters reported that Trump was expected to sign an order on AI and cybersecurity as soon as Thursday and that White House officials were trying to assemble AI chief executives for a signing ceremony. Federal News Network reported earlier this month that National Economic Council Director Kevin Hassett compared the concept to pre-release safety testing, saying the administration was considering ways to ensure new models are secure before they are released publicly. Politico reported on May 5 that the administration had been discussing executive actions on frontier models, including a vetting regime focused on national security risks. (kesq.com) ### Why did Anthropic’s Mythos become the trigger? Anthropic’s Mythos became a focal point after the company said the model could exploit cybersecurity vulnerabilities at unprecedented speed, according to CNN’s report. Reuters reported on May 18 that Anthropic had revised its earlier position to allow Mythos users to share cyber-threat information with others who might face similar vulnerabilities. (federalnewsnetwork.com) Anthropic has limited access to Mythos through a controlled group of partners, according to reporting summarized in recent coverage and the company’s own broader cybersecurity initiative. Anthropic said in a separate announcement on Project Glasswing that it was bringing together companies including Amazon Web Services, Apple, Cisco, CrowdStrike, Google, Microsoft and Palo Alto Networks to secure critical software in what it called the AI-driven era of cybersecurity. (kesq.com) ### What rules is OpenAI trying to put in place? Politico reported on May 20 that OpenAI is backing a framework centered on transparency and reporting requirements for developers of advanced AI. Politico said those proposals would give OpenAI a more stable legal framework while creating relatively little new liability for catastrophic harms compared with stricter safety proposals. (anthropic.com) A separate Politico report said OpenAI has also been pursuing a state-by-state lobbying strategy as legislatures across the United States consider their own AI bills. That campaign has unfolded as House lawmakers and White House officials weigh whether some state AI laws should be blocked or preempted, according to Politico’s May 15 reporting. ### Why are companies fighting in the states as Washington moves? State legislatures have become a second front because Congress has not produced a national AI statute and states have kept advancing their own proposals. (politico.com) Politico reported that OpenAI’s lobbying push is aimed at shaping those state rules before stricter approaches gain traction. California already enacted a transparency law for frontier AI developers last year, according to legal analysis from Crowell & Moring describing the state measure. (politico.com) That law requires large developers to publish a safety framework and explain how they address catastrophic risks, adding pressure for companies to influence how other states write their own bills. (politico.com) ### What happens next at the White House? Thursday, May 21, is the earliest reported date for Trump to sign the order, according to Reuters and CNN. Reuters said White House officials were working to get AI company CEOs to Washington for a signing ceremony, while CNN said the order would center on voluntary pre-release sharing of advanced models with the government. Whether that ceremony happens on Thursday or later, the next public marker is expected to be the text of the executive order and the list of companies willing to join the voluntary review process. (crowell.com) (kesq.com)