White House Condition for Anthropic’s Fable 5: The Trump administration has moved to limit access to Anthropic’s most advanced AI models outside the United States, and that has put the company under fresh pressure from Washington. According to recent reporting, officials are not ready to loosen the rules unless Anthropic can answer fears about “jailbreaks” and other security holes.
Anthropic says the government ordered it to suspend access to Fable 5 and Mythos 5 for foreign nationals, whether they are inside or outside the U.S., and the company says it removed access for all customers to stay in line with that order.
House of the Dragon Season 3 Set to Unleash the Deadliest Chapter of the Dance of the Dragons
Why the White House is Worried
The concern is not just about a chatbot saying the wrong thing. Wired reports that White House officials asked the NSA to review the model’s weak spots, and the agency concluded that it was possible to strip away Fable 5’s guardrails. Those guardrails were meant to block dangerous help in areas like cybersecurity, biology, and chemistry. Anthropic itself has also said that large language models remain open to jailbreaks and that no AI system on the market has perfectly robust defenses.
In Anthropic’s own words, the government told it to suspend access because of national security concerns. The company said it had only been given verbal evidence of a “potential narrow, non-universal jailbreak” and argued that this was not enough reason to pull a commercial model that had already been used at scale. That disagreement sits at the heart of the fight now.
Why Other Leaders are Jumping in?
The dispute has also spilled into wider global AI politics. Reuters reported that the European Commission is still talking with Anthropic about the decision to disable its most advanced models in the EU.
At the G7 summit in France, leaders and tech executives discussed whether trusted partners should still get access to frontier AI systems. French President Emmanuel Macron also pushed for more cooperation among democracies instead of a split world where powerful AI tools are kept behind national walls.
When Salary Isn’t Enough: Why Financial Preparedness Is Becoming a Survival Skill for Young Indians
Why Experts don’t see an Easy Fix
Even with stronger safety systems, the problem is messy. Anthropic says its own safety work is focused on jailbreak resistance, including special defenses against universal jailbreaks. But the company also admits that no current AI system is perfectly sealed. That is why some researchers think the government is asking for something almost impossible: a model that can never be tricked by a determined user.
