Negotiations between Anthropic and the White House over the restriction on Fable 5 and Mythos 5 are reportedly moving more smoothly after CEO Dario Amodei stepped back from direct involvement. The shift has become a notable turning point in talks that had shown little progress despite continued efforts from both sides.
According to Wired, the discussion is now being led by Anthropic co-founder Tom Brown and public policy chief Sarah Heck. Government officials are said to feel more comfortable engaging in the talks since Amodei’s role at the table has been reduced.
A different style of negotiation
The change in personnel appears to matter as much as the substance of the talks. Officials reportedly viewed Amodei as difficult to work with and said he was not listening closely to their concerns.
One source familiar with the calls described Brown as someone who can genuinely participate in the discussion, unlike Amodei. That perception has helped improve the tone of the exchanges, even though no final decision has been announced.
The ban itself remains in place. The Trump administration and Anthropic have continued holding several calls in recent days, including technical discussions about what evidence the company would need to provide to address the government’s concerns.
What triggered the dispute
The dispute began after the U.S. government directed Anthropic to block foreign nationals, including its own employees, from accessing Mythos 5 and Fable 5. Anthropic chose to remove access to the two models for all users instead of limiting the restriction only to the groups named in the original directive.
That rapid move was initially praised by Trump. In comments to Axios, Trump said Amodei responded very quickly because the issue was a major responsibility, and he called Anthropic’s reaction responsible.
| Model | Issue Raised | Status |
|---|---|---|
| Claude Fable 5 | Jailbreak concerns and possible weakened safeguards | Access still restricted |
| Claude Mythos 5 | Foreign access concerns and broader security scrutiny | Access still restricted |
Why Fable 5 and Mythos 5 are under scrutiny
The disagreement centers on Claude Fable 5 and Claude Mythos 5, both part of Anthropic’s “Mythos-class” family. The company says those models are built on the same core technology as Claude Mythos Preview, which had previously been considered too dangerous to release publicly.
Fable 5 was introduced as a version with safeguards meant to reduce cyber risk. Soon after launch, however, concerns emerged that the protection layer could be bypassed.
Wired reported that researchers at Amazon alerted White House officials shortly after Fable 5’s release that jailbreak techniques could remove part of its protection. Before the jailbreak issue surfaced, the White House was also said to worry that parties linked to Chinese interests may have accessed the Mythos-class models.
Those concerns now shape the ongoing talks between the government and Anthropic. A major point of discussion is what standard of proof would be enough to convince officials that the jailbreak problem on Fable 5 has been handled adequately.
Technical review continues in parallel
The negotiations are not limited to senior executives. The report says the two sides have also been meeting through a working group that includes technical staff from both companies.
This format suggests the dispute is not only about corporate relationships, but also about evidence and verification. The government wants to know what data or testing can demonstrate that the model protections are strong enough to prevent misuse.
Despite the more constructive tone, there is still no sign that access to Fable 5 and Mythos 5 will be restored soon. The restriction remains active while the technical and policy questions continue to be worked through.
The broader debate over AI guardrails also hangs over the talks. Independent cybersecurity experts have increasingly argued that safeguards on AI models are only temporary, because highly skilled users and newer AI systems can still find ways around them.
That debate complicates Anthropic’s position further. Even if the company can show stronger protection, the larger question of how far guardrails can be trusted remains unresolved.
Not the first friction point
This is also not the first time Anthropic has clashed with the U.S. government. Earlier this year, the AI startup was labeled a supply chain risk by the Department of Defense after a dispute over unrestricted AI use.
That history provides context for why the current discussions are being watched closely. For the government, the issue touches national security, foreign access, and the potential misuse of advanced models.
For Anthropic, the outcome will determine whether Mythos-class models can return to access and under what conditions. For now, the talks are said to be on a better footing with Tom Brown and Sarah Heck leading the discussions, but the White House has not yet issued a final decision on lifting the ban.
