Dean Ball is running headfirst into the future. One of the most visible experts in the fledgling world of artificial intelligence policy, he spent much of 2025 advising the White House on emerging technologies before moving to the Foundation for American Innovation. Then, just yesterday, he announced he was joining OpenAI, to lead the new “Strategic Futures” team. “An incredible opportunity that came at the right time!” he texted me.
Ball sounded a bit frantic when I called him on Friday, June 12, around 3 p.m. We were supposed to meet for drinks an hour later, but I hadn’t gotten a confirmation. Something came up, he said on the phone, and he’d been dealing with it. We arranged to meet at a hotel bar in Midtown Manhattan.
He didn’t say what had been keeping him busy. But around the time we were ordering our second round of drinks—5:21 p.m. on Friday—the AI company Anthropic received a letter from the Department of Commerce informing them that two of their newest and most powerful AI models, Claude Fable 5 and Mythos 5, had been placed under an export control. According to Anthropic, this meant that they were banned from allowing foreign nationals, whether inside or outside the U.S., to access either product. “The net effect of this order is that we must abruptly disable Fable 5 and Mythos 5 for all our customers to ensure compliance,” the company wrote.
This is the same model that, Anthropic said, “found thousands of high-severity vulnerabilities,” when it was first released to a limited number of companies in April, “including some in every major operating system and web browser.” Reportedly, the order came after Amazon CEO Andy Jassy told White House officials that his team had found a way to “jailbreak” Fable 5—which is, confusingly, a version of Mythos—by getting the model to give them classified information about cyber attacks.
