Will Anthropic fix its Fable jailbreak to lift US
Published 6/14/2026, 4:42:40 AM
Answer
The situation remains unresolved, and the premise of the question is contested by Anthropic and independent cybersecurity experts.
What the Evidence Shows
Anthropic disputes the "jailbreak" characterization entirely. After examining the published examples, Anthropic determined that some outputs were not produced by Fable 5 at all, and those that were contained only general information already available in public sources. They state: "No evidence of safeguards being successfully circumvented to generate genuinely dangerous content."
Independent experts agree. Katie Moussouris, CEO of Luta Security, reviewed the research and stated publicly: "I've seen the paper. It's not a jailbreak. It was Defense Oriented Prompting (DOP), capabilities defenders need." She criticized the government's action as an "own goal against us" for national defense.
The US government issued a directive on June 12, 2026 requiring Anthropic to suspend access to Fable 5 and Mythos 5 for foreign nationals. However, Anthropic notes the government provided only "verbal evidence" with no specific details about the national security concern.
Why the Question's Premise Doesn't Hold
| Claim | Status | Evidence Gap |
|---|---|---|
| A true "Fable jailbreak" vulnerability exists | Contested | Anthropic and cybersecurity experts say it is not a jailbreak |
| Anthropic committed to fixing it | Not supported | Anthropic is arguing it is a misunderstanding, not a defect requiring a fix |
| Fixing it would lift export controls | Unclear | The causal chain is not established in available evidence |
Anthropic's official position is: "We believe this is a misunderstanding and are working to restore access as soon as possible." They are working to restore access by arguing to the government that the technique is legitimate defensive capability, not by committing to patch a vulnerability.
What Remains Open
- Whether the government will accept Anthropic's framing that this is "Defense Oriented Prompting" rather than a true jailbreak
- Whether the same standard applied to OpenAI's GPT-5.5 (which Anthropic notes has the same capability "widely available")
- Whether the export control will be lifted without a technical fix, or whether Anthropic will eventually implement one under pressure
Bottom line: Anthropic has not committed to fixing a "jailbreak" because they do not believe one exists. The path to lifting export controls depends on whether the government accepts their argument that the technique is a legitimate defensive tool, not a vulnerability requiring remediation.