Anthropic Adds Guardrail to Claude Fable 5 After Export Control Dispute With Trump Administration
Anthropic agreed to extend security safeguards on Claude Fable 5 to resolve export restrictions imposed by the Commerce Department.
Last verified:
Export Control Dispute Resolved Through Guardrail Extension
Anthropic has secured the lifting of export controls on Claude Fable 5 and Mythos 5 by agreeing to extend a security safeguard that automatically downgrades restricted capability requests to a less-capable model, according to two people familiar with the matter cited by Wired AI. Under the new measure, users attempting to access sensitive cybersecurity or biology features will be notified their request is blocked and their query will be routed to the Opus 4.8 model instead. The deal clears a significant regulatory hurdle for the company, though challenges with other administration agencies remain unresolved.
The Workaround That Sparked the Impasse
The conflict began when researchers identified a technique to circumvent Claude Fable 5’s existing guardrails on sensitive capabilities. According to an analysis by Katie Moussouris, founder and CEO of Luta Security, users could request that the model fix code rather than identify security issues—effectively bypassing the intended restriction. Wired AI reports that cybersecurity experts generally did not consider this behavior a genuine security risk, but the Trump administration’s Commerce Department escalated the issue by imposing export controls that rendered Fable 5 commercially inaccessible.
The administration had previously implemented guardrails on Fable 5 that processed sensitive cybersecurity and biology requests through the less-advanced Opus 4.8 model. Anthropic’s new safeguard extends this filtering logic to the specific indirect technique outlined in an Amazon research paper, closing what the administration viewed as a loophole.
Commerce Department Approval and Defense Department Holdout
Commerce Secretary Howard Lutnick announced the removal of export restrictions, stating in a letter obtained by Wired AI that “Anthropic has agreed to proactively detect and address security risks posed by the models.” The Commerce Department’s Center for AI Standards and Innovation determined that Fable 5’s updated safeguards were sufficiently robust to permit public release.
However, Anthropic’s regulatory challenges are not fully resolved. According to sources briefed on the matter, Defense Secretary Pete Hegseth has declined to lift a February 28 order classifying Anthropic as a supply-chain risk, with no apparent pathway toward removal of that designation. This distinction matters: while export controls have been lifted, the supply-chain designation may constrain government procurement and partnerships.
Why This Matters
Anthropic’s concession illustrates a narrower pattern in tech regulation during the Trump administration: targeted security guardrails on specific capability vulnerabilities, rather than blanket export prohibitions on model classes. For developers and enterprises deploying Claude Fable 5, the lifting of export controls restores access and unifies the international market. However, the unresolved Defense Department designation signals that regulatory risk remains for Anthropic and similar companies operating in sensitive capability spaces. The precedent also establishes that administration officials may respond to novel jailbreak techniques published in academic research by imposing trade restrictions—a pattern that could incentivize either preemptive disclosure of vulnerabilities or defensive publication strategies among AI labs.
Frequently Asked Questions
What specific guardrail did Anthropic add to Claude Fable 5?
Anthropic extended an existing safeguard to detect attempts to access restricted capabilities—particularly those identified in an Amazon paper on indirect code-manipulation techniques. Blocked requests are now automatically processed by Opus 4.8 instead.
Why did the Trump administration impose export controls on Anthropic in the first place?
According to Wired AI, the administration learned about a technique where users could circumvent Fable 5's security restrictions by asking the model to fix code rather than identify security issues. While cybersecurity experts did not view this as a genuine threat, the administration used it as justification for export controls.
Is Anthropic fully back in compliance with the administration?
Partially. The Commerce Department lifted export restrictions on Fable 5 and Mythos 5 after Anthropic's concessions. However, Defense Secretary Pete Hegseth has not lifted a February 28 order designating Anthropic a supply-chain risk, and sources say no clear path exists to do so.