US Locks Down Anthropic AI Over "Jailbreak" Warning

The US government has directed Anthropic to suspend access to its Fable 5 and Mythos 5 AI models for all foreign nationals. This action was taken due to national security concerns. Anthropic blocked access to these models on Friday evening. The government believes a method exists to bypass, or "jailbreak," Fable 5. The company is now on the Pentagon's blacklist. US Commerce Secretary Howard Lutnick sent a letter to Anthropic's CEO, stating the models should remain locked down for a few weeks. In the future, Anthropic will need a license for the transfer of these models. The concern is that these AI models could pose a catastrophic risk to public and national security. They might find loopholes in software, create weaponized exploits, or even disclose information for dangerous pathogens. Jailbreaking refers to bypassing the safeguards Anthropic put in place to prevent misuse. Anthropic says demonstrations of this technique only identified minor vulnerabilities, and other public models can also find them. They also state no tester has found a "universal jailbreak." The bottom line is that the US government is taking significant steps to control access to powerful AI, highlighting growing concerns about its potential risks.