US Bans AI Export for First Time, Signaling Start of Digital Cold War

President Donald Trump has blocked the export of a powerful artificial intelligence model for the first time, marking a new phase in the global tech race. The decision forces the startup Anthropic to cut all foreign access to its advanced AI system, called Mythos. Mythos is a cutting-edge AI model that can find security flaws in computer systems—weaknesses that even the best human experts missed. This makes it a potential weapon for cyberattacks on banks, governments, and critical infrastructure. Until now, U.S. controls focused on limiting chip sales to China. This new move goes much further. It blocks anyone outside the U.S., including foreign workers at Anthropic, from using the AI. The order came suddenly on a Friday evening. Anthropic says it received the directive at 5:21 PM with no detailed explanation of the specific security concern. **The "Beautiful Baby"** Trump previously took a hands-off approach to AI regulation. At the start of his term, he called AI a "beautiful baby" and said, "We cannot stop it with politics. We cannot stop it with stupid rules." He appointed two Silicon Valley insiders—Elon Musk and investor David Sacks—as special government employees. They pushed for no regulation. Sacks argued that AI rules would hurt the U.S. in its race against China. Trump initially canceled all AI safety measures put in place by President Joe Biden. He also tightened chip sales to China. **The Mythos Trigger** Anthropic, founded by former OpenAI employees who prioritize safety, released Mythos in April. It did not make the AI public. Instead, it gave limited access to big tech companies, banks, and governments for security testing. Banks like JP Morgan grew worried. They told Trump that if Mythos leaked, it could enable massive cyberattacks. **The Policy Reversal** Treasury Secretary Scott Bessent and Chief of Staff Susie Wiles convinced Trump that doing nothing would make the government look responsible for a future AI-powered cyberattack. Trump planned to sign an order requiring companies to submit new AI models for government review 90 days before release. But on the morning of May 21, he hesitated after calls from Musk, Sacks, and Meta CEO Mark Zuckerberg. "I didn't like certain aspects," Trump said. "I think it gets in the way... we are leading China. I don't want to do anything that gets in the way." He signed a weaker order on June 2. It required only a voluntary 30-day notice for advanced AI models. **The Final Ban** Ten days later, Anthropic released Fable 5, a less powerful version of Mythos. The White House suspected it had been hacked. Anthropic says it found no evidence of a security breach. Despite this, Trump ordered a complete ban on foreign access. The decision also blocks Spain’s planned access to Mythos, which was granted earlier this month for cybersecurity testing. Spanish officials say it is unclear what will happen next. This is the first time the U.S. has blocked an AI model already used by millions worldwide. Experts say it may not be the last.