Anthropic Restores Claude Fable 5 After Export

What happened to Claude Fable 5 and Mythos 5?

Anthropic restored access to Claude Fable 5 and Claude Mythos 5 on July 1, 2026, after the US government lifted export controls it had imposed on June 12. The suspension lasted 18 days and affected all users globally. Because the original order took effect immediately and Anthropic had no reliable way to verify nationality in real time, the company suspended both models for everyone — not just foreign nationals.

The export control directive came after Amazon researchers found a method of bypassing Fable 5's safety classifiers. The technique prompted the model to identify software vulnerabilities. In one case, the model produced code showing how a specific vulnerability could be exploited, according to Anthropic's redeployment statement.

What is Claude Fable 5?

Claude Fable 5 is Anthropic's most capable widely released model, built for demanding reasoning and long-horizon agentic work. Claude Mythos 5 shares the same underlying model but is available only in limited release through Project Glasswing, a restricted program for trusted partners. Fable 5 includes safety classifiers; Mythos 5 does not.

What did the Amazon researchers actually find?

Amazon researchers found a technique that bypassed Fable 5's safety classifiers, allowing the model to identify software vulnerabilities and, in one instance, produce code demonstrating how to exploit a single vulnerability.

Anthropic tested the same technique against multiple other models. Every model tested — including Claude Haiku 4.5, Sonnet 4.6, Opus 4.6, Opus 4.7, Opus 4.8, GPT-5.4, GPT-5.5, and Kimi K2.7 — could produce the same demonstration. The behavior did not expose any unique Mythos-level cyber capabilities. Anthropic described it as "a borderline case" for Fable 5's classifiers involving "routine defensive cybersecurity work."

You might also like

RoguePlanet CVE-2026-50656: Defender Zero-Day

OpenAI Releases GPT-5.6 Sol & National Security

Anthropic Fable 5 & Mythos 5 Export Ban Lifted

Prince George's County Passes 2-Year Data Center

How did Anthropic fix the safety issue?

Anthropic trained a new safety classifier that specifically targets and blocks the behavior described in the Amazon report. The new classifier blocks the reported technique in over 99% of cases. In a very small fraction of cases, the model may provide information that isn't detailed enough to assist a cyberattacker.

Researchers from the US Department of Commerce's Center for AI Standards and Innovation (CAISI) tested both the prior and new safeguards and agreed they are "extraordinarily strong," per Anthropic's statement.

The fix comes with a trade-off. The new classifier flags benign requests more often during routine coding and debugging tasks. Anthropic said it will continue refining the classifier to reduce false positives.

What changed in the API for developers?

The key developer-facing change is that Claude Fable 5 now includes safety classifiers that can decline requests. Claude Mythos 5 does not include these classifiers.

When Fable 5 declines a request, the Messages API returns stop_reason: "refusal" as a successful HTTP 200 response — not an error. The response also reports which classifier declined the request. Developers integrating Fable 5 need to plan for three changes, per the Claude platform documentation:

• New response handling for refusals (HTTP 200, not an error)
• Fallback options for retrying on another Claude model
• New billing rules that apply specifically to refusals

Where is Fable 5 available now?

For Pro, Max, Team, and select Enterprise plans, Fable 5 is included for up to 50% of weekly usage limits through July 7. After that date, it will be available via usage credits.

What is Project Glasswing?

Project Glasswing is Anthropic's restricted program that gives a limited set of trusted partners access to Claude Mythos 5. As of July 1, Mythos 5 access was restored only for US organizations that received government approval on June 26. Anthropic said it continues to coordinate with the government to expand access to the broader set of domestic and international Glasswing partners.

What is the industry framework Anthropic is proposing?

The suspension made clear to Anthropic that the industry needs a consistent way to assess and fix AI model jailbreaks — techniques that bypass a model's safeguards. Anthropic said it has started developing a shared standard with Amazon, Microsoft, Google, and other Glasswing partners. The goal is a framework that helps AI developers triage new findings, launch capable models more safely, and communicate risk levels consistently to government and industry.

Here's what we know so far: the framework is still in development, and Anthropic has not published a final version.

What is Anthropic's broader business position?

The SemiAnalysis newsletter reported that Anthropic confidentially filed for an IPO on June 1, 2026. The newsletter described Anthropic as "the clear clubhouse leader in capturing the B2B market" in 2026, citing Claude Code's traction in software development. Anthropic and OpenAI together account for roughly $100 billion in ARR, according to that SemiAnalysis report. The newsletter also noted that Dario Amodei left OpenAI to start Anthropic in early 2021, over 18 months before ChatGPT's viral release.

Anthropic's NYC office expansion and Amazon's $25B bond sale for AI infrastructure spending both reflect the scale of investment flowing into this space. The SpaceX Nasdaq 100 inclusion on July 7 is another marker of how AI-adjacent companies are reshaping public markets at the same moment Anthropic is preparing its own IPO filing.

The next confirmed milestone: Fable 5 transitions from the 50%-of-weekly-usage-limits offer to usage-credits-only access on July 7, 2026.