On June 30, 2026 — 18 days after the Commerce Department ordered Anthropic to pull it Commerce Secretary Howard Lutnick sent a letter to ...
The model has a new safety classifier trained specifically to block the jailbreak that triggered the ban in the first place. Billing has changed usage credits replace the free-access period that was originally promised. Anthropic has signed a formal cooperation agreement with the US government covering protocols for future model releases, threat reporting, and malicious use disclosure. The return of Fable 5 is real, but it comes with a changed relationship between Anthropic and the federal government that will shape how every future frontier model gets released.
What Triggered the Ban and What Anthropic Found When They Investigated
Fable 5 and Mythos 5 launched on June 9. Both share the same underlying model Fable 5 was the version released publicly with stronger safety guardrails, Mythos 5 the version with fewer guardrails, limited to trusted Glasswing program partners for defensive cybersecurity work. Three days after launch, on June 12, the export control directive arrived. The trigger was a report from Amazon researchers demonstrating a jailbreak: a specific prompting technique that got Fable 5 to identify software vulnerabilities and, in one case, write code showing how one could be exploited.
Anthropic spent the following 18 days working with the government, Amazon, and other partners to review the report and test the finding. Their conclusion, published on June 30, is notable: testing showed that many less capable models including Claude Opus 4.8, GPT-5.5, and Kimi K2.7 could identify the same vulnerabilities Fable 5 did. In other words, the specific technique that triggered the export control worked on models already widely available. That finding shaped how Anthropic argued for the ban's removal. If the capability is already present in models that weren't banned, the export control on Fable 5 specifically wasn't containing the risk it was designed to address.
Context matters here too. Earlier this spring, Anthropic had tested a prior Mythos model that found and exploited zero-day bugs across every major operating system and browser on command including a 27-year-old flaw in OpenBSD. That capability is what the government is actually worried about. Fable 5's jailbreak produced something meaningfully less severe than that, which is part of why the ban was eventually lifted rather than extended.
What the New Safety Classifier Does and What It Costs You
As a condition of return, Anthropic trained a new AI-based safety classifier specifically designed to block the jailbreak technique identified in the Amazon report. According to Anthropic's June 30 writeup, the classifier now stops that technique in more than 99% of attempts. When a blocked request is detected, Fable 5 falls back to Opus 4.8 instead, and the user is notified.
The trade-off Anthropic acknowledges directly: more false alarms on normal coding and debugging requests. The classifier is trained on the specific jailbreak technique, which means it will occasionally flag legitimate vulnerability research or security-adjacent coding tasks as potentially problematic. Anthropic says this is the right trade-off given the severity of the risk category, but enterprise security teams who use Fable 5 for legitimate code review and vulnerability assessment should expect some operational friction that wasn't present in the original June 9 release.
Anthropic also opened a HackerOne bug bounty program specifically for Fable 5 jailbreak reports, and committed to deploying fixes within hours for the most severe jailbreak categories specifically those enabling attacks on power grids or financial systems. A 24/7 jailbreak monitoring team has been stood up. For the worst-case scenarios, the response model is now immediate rather than reactive.
How Billing Has Changed What You Need to Know Before July 7
When Fable 5 originally launched on June 9, it was free through June 22 for all users. That arrangement is gone. The new billing structure depends on your plan type and changes on July 7.
For Pro, Max, and Team plan subscribers: through July 7, Fable 5 is included in your subscription for up to 50% of your weekly usage limits. After July 7, Fable 5 moves to usage credits only billed at the same rates as API access. If you don't have usage credits enabled, Fable 5 will stop working for your account after July 7.
For Enterprise customers, the situation splits by seat type. Standard Enterprise seats get no included Fable 5 allowance at any point all usage is billed through credits from day one. If credits aren't enabled, Fable 5 doesn't work. Premium Enterprise seats get a short grace period through July 7 where Fable 5 draws from each member's seat usage at no additional cost then moves to credits. API pricing remains unchanged at $10 per million input tokens and $50 per million output tokens. AWS, Google Cloud, and Microsoft Foundry re-enablement is coming but has no confirmed date yet.
What Anthropic Agreed to as a Condition of Return
Lutnick's letter to Tom Brown specified what Anthropic agreed to in exchange for the export controls being lifted. The commitments, confirmed by NBC News and CoinDesk: continuing collaboration with the government on protocols and standards for Mythos, Fable, and future models; informing the US government of any malicious activity detected; and giving designated government agencies early access to frontier models and their safeguards before public release.
Anthropic is also working with Amazon, Microsoft, and Google to develop a common framework for scoring how dangerous a given jailbreak is. That's a piece of infrastructure that doesn't currently exist in the industry a shared severity classification system that would give both companies and regulators a consistent vocabulary for talking about what a given jailbreak actually enables.
The Hacker News analysis puts the underlying situation plainly: "When Washington wants to move fast on a frontier model, it still has no binding process, only improvised ones." The export control on Fable 5 was applied under trade law, not AI regulation. The June 2 executive order created a voluntary path for pre-release review. None of it is legally binding in the way that, say, FDA drug approval or aviation safety certification is. What happened with Fable 5 was a government improvising with the tools it had, and Anthropic negotiating its way out of a ban that was built on the same improvised foundation. The cooperation agreement formalized that relationship which may ultimately be more significant than the ban itself.
Frequently Asked Questions
Q: When does Fable 5 return and on which platforms?
Fable 5 returns globally on July 1, 2026 across Claude.ai, the Claude Platform, Claude Code, and Claude Cowork for Pro, Max, Team, and Enterprise plans. AWS, Google Cloud, and Microsoft Foundry re-enablement is coming but has no confirmed date.
Q: What changed in Fable 5 after the export control ban?
Anthropic added a new safety classifier trained to block the Amazon-discovered jailbreak technique stopping it in more than 99% of attempts. Blocked requests fall back to Opus 4.8. The trade-off is more false positives on legitimate coding and security tasks.
Q: How does Fable 5 billing work after July 7, 2026?
After July 7, all plans move to usage credits. Pro, Max, and Team get up to 50% of weekly limits included through July 7 then need credits enabled. Standard Enterprise has no included allowance credits required from day one. Premium Enterprise gets the grace period through July 7 then credits. API pricing stays at $10 per million input tokens and $50 per million output tokens.
Q: What did Anthropic agree to with the US government?
Anthropic agreed to give government agencies early access to frontier models before release, report malicious activity, collaborate on future release protocols, and work with Amazon, Microsoft, and Google on a shared jailbreak severity scoring framework. These commitments were confirmed in Commerce Secretary Lutnick's letter to Tom Brown on June 30.
References
- Anthropic. Redeploying Claude Fable 5. June 30, 2026. anthropic.com
- NBC News. Commerce Department gives green light for Anthropic to bring back Fable 5. June 30, 2026. nbcnews.com
- The Hacker News. Anthropic Restores Claude Fable 5 After US Lifts Jailbreak-Linked Export Controls. July 1, 2026. thehackernews.com
- The New Stack. How Anthropic is bringing Fable 5 back and when it'll cost you. July 1, 2026. thenewstack.io
