US bans Anthropic's newest models, researchers call the move dangerous

The US government forced Anthropic to pull its two newest models, Fable 5 and Mythos 5, late last week, citing national security concerns. The trigger: Amazon researchers reportedly found a way to bypass Fable 5's guardrails. The move is already drawing pushback from the people who study this for a living.

Cybersecurity researchers have signed an open letter calling the ban dangerous. Anthropic itself noted that the same jailbreaks the government cited exist in other models that remain live. According to TechCrunch, the inconsistency is raising questions about whether the action was driven by methodology or something else.

The broader question is whether selective enforcement on one model while ignoring parallel vulnerabilities in others actually improves security or just fragments the conversation. Anthropic has not said whether it plans to re-release the models with updated guardrails, and the government has not clarified what threshold would allow their return.

Meanwhile, some observers are asking whether the ban inadvertently strengthens Anthropic's brand. Being singled out by the US government for building models advanced enough to warrant a national security review is a different market signal than being ignored. The company has not commented on that angle, but the perception is already circulating in AI-adjacent circles.

The immediate impact: two models are off the market, the research community is pushing back, and the regulatory precedent is unclear. What happens next depends on whether the government defines a consistent standard or whether this remains a one-off intervention.