Trump signs executive order letting AI companies submit models for government review

President Trump signed an executive order creating a voluntary cooperation program for AI companies. Before releasing major AI models, companies can choose to submit them to the federal government for review. The goal is to boost innovation and strengthen cybersecurity for key infrastructure. The order says the U.S. AI industry thrives partly because it avoids heavy regulations, but it also admits new AI tech comes with safety risks. Federal agencies will set rules to check models' cyber attack and defense abilities before they go public.

Companies can decide if they want to hand over their models before launch. Those that do get some privacy protections. Trump had delayed signing an earlier version because he worried it would hurt U.S. competition with China. The new rule shortens the review window to at most 30 days before release. Google, Microsoft, and xAI have already agreed to let a Commerce Department group check their models for safety.

Back in 2024 under Biden, OpenAI and Anthropic had already signed up to submit their models. After taking office, Trump's team played down AI safety risks at first, but this order shows a shift. One big reason is Anthropic's new model Mythos, which found thousands of serious security holes by itself. Industry groups support the new rule and are asking Congress to make these safety measures a legal requirement.