Anthropic made Claude Fable 5 available again to users worldwide on 1 July 2026, two days after the United States government lifted the export controls (rules that restrict who can access a given technology) that had forced the company to suspend access to the model on 12 June. The return was announced by the company and confirmed by several news outlets.
Fable 5 is the most capable artificial intelligence model Anthropic has released to the public, launched on 9 June. Three days later, on 12 June, the US Department of Commerce imposed export controls that required restricting access for foreign nationals, inside or outside the country. According to Anthropic, because the order took effect immediately and there was no reliable way to verify nationality in real time, the company chose to suspend access for all users.
Why it was pulled and why it is back
The directive followed a report in which Amazon researchers described a way to bypass Fable 5's safety guardrails (a technique the industry calls a jailbreak): getting the model to identify software flaws and, in one case, to produce code demonstrating how one of those flaws could be exploited.
In subsequent testing, Anthropic says it confirmed that several less capable models could identify the same vulnerabilities, among them its own Opus 4.8, OpenAI's GPT-5.5 and Kimi K2.7. For the exploit demonstration, it says every model tested reached the same result. The company concludes that the technique exposed no offensive capabilities unique to Fable 5 and amounted to routine defensive cybersecurity work.
Even so, it trained a new classifier (an automated system that detects and blocks requests deemed dangerous) targeting that behavior. According to the company, the classifier now blocks the technique described in the report in more than 99% of cases, and blocked requests are answered by the next model down, Opus 4.8. CAISI, the AI standards center at the Department of Commerce, tested the safeguards and, in Anthropic's words, found them "extraordinarily strong." The export controls were lifted on 30 June.
Where and at what price it returns
Fable 5 is available again globally on the Anthropic platform, on Claude.ai, Claude Code and Claude Cowork, with access on Amazon (AWS), Google Cloud and Microsoft Foundry being restored "as quickly as possible," the company says. Pricing stays at $10 per million input tokens and $50 per million output tokens (tokens are the units of text the models process); through the programming interface (the API), access is billed at those rates.
On subscription plans, the company opened a promotional window between 1 and 7 July, during which Pro, Max, Team and some Enterprise seats can use Fable 5 for up to 50% of their weekly limits at no extra cost. After that, use requires usage credits.
Access to Mythos 5, the model with the same capabilities but fewer safeguards, was restored only for a set of US organizations, following the US government's approval on 26 June. It remains reserved for partners in the Glasswing program.
A common yardstick for "jailbreaks," and a bigger state role
Alongside the return, Anthropic announced it is working with Amazon, Microsoft, Google and other Glasswing partners to build a common framework for assessing the severity of jailbreaks, with four criteria: the capability gain the technique gives the attacker, its breadth, how easily it can be turned into an attack, and how easily it can be obtained. It also launched a HackerOne program for researchers to submit flaws they find in Fable 5.
The resolution also involved deeper collaboration with the US government, under the 2 June executive order (a presidential decree) on AI innovation and security. Anthropic committed to giving government partners early access to frontier models and their safeguards, and to sharing information about jailbreaks and misuse. According to Euronews, the lifting was communicated by the Department of Commerce; CNBC and VentureBeat reported that Commerce Secretary Howard Lutnick authorized the reversal after two weeks reviewing the models with the company.
Anthropic itself acknowledges that the new classifier flags more legitimate requests by mistake during routine coding tasks, and says it will keep tuning the safeguards to reduce those false positives. Still to come are the wider expansion of Mythos 5 access and the finalization of the proposed severity framework, which the company describes as a work in progress.
Sources: Anthropic (Redeploying Fable 5), Anthropic (Claude Fable 5 and Mythos 5), promotional access, Euronews.
#StaySafe
🙏🖖
