The US government instructed Anthropic to disable its two newest and most powerful AI models, Mythos 5 and Fable 5, just three days after their launch. The directive stated that no one outside the US could use these models. Since Anthropic was unable to isolate US users from international users quickly, they decided to shut down both models for everyone, everywhere.
The government claimed someone found a way to trick Fable 5 into ignoring its safety rules (a “jailbreak”). However, Anthropic responded that the evidence was weak, noting that it involved only a person asking the AI to analyze code and fix bugs, and that other AI models can do the same anyway.
Anthropic had already made the government upset months earlier. Anthropic refused to let the military use Claude for spying on US citizens or for autonomous weapons that could operate without a human. The government didn’t like that, so things escalated from there. Trump even said other branches of the US government need to stop using Anthropic within 6 months.
A few government offices have already switched to other AI tools instead of Claude. Some businesses got disrupted too and had to switch to a different Claude model in the meantime.
This is just about two specific, brand-new AI models, not Claude as a whole. The Claude you’re chatting with right now is totally unaffected.
