Anthropic built a coding-focused AI model called Mythos, deemed dangerous enough to share only with a small group of cybersecurity experts. Its modified release, Fable, went public on June 9. Four days later, the White House declared it a national security threat and placed export controls on it. Anthropic pulled both models within hours. The trigger: Amazon CEO Andy Jassy, whose company is both an Anthropic investor and a competing AI builder, told government officials Fable was dangerous.
Three consequences are already unfolding. First, European leaders like Bruno Retailleau are calling this a wake-up call to build sovereign AI, but Chinese open-source models from companies like Zhipu are cheap, capable, and downloadable with no guardrails, making them attractive to any company that does not want the White House deciding their access. Second, the ban may be making the US more vulnerable, not less. An open letter at freefable.org signed by leading cybersecurity researchers argues Anthropic's models were actively helping defenders, and that Fable is no more dangerous than widely available alternatives. Third, Congress is watching. After Anthropic's last government clash over Pentagon usage, a slate of new military AI bills followed. The same escalation is possible now.
The full piece is worth reading for what it does not resolve: whether the export control logic even applies to software, whether a legal challenge will kill the ban, and what happens if US companies decide Chinese models are simply easier to work with. The administration that scrapped AI safety rules on day one has now called the same startup a national security risk twice in one year. The pattern matters more than any single decision.
[READ ORIGINAL →]