AI
Claude Fable 5 Returns Globally Tomorrow, With Tighter Cybersecurity Guardrails
Anthropic will restore global access to Claude Fable 5 starting July 1, ending a nearly three-week outage triggered by a U.S. export control directive.
3 min read
By Timmy
Anthropic will restore global access to Claude Fable 5 starting July 1, ending a nearly three-week outage triggered by a U.S. export control directive. The model returns with a new layer of classifiers aimed at blocking cybersecurity misuse — a shift that will, at least temporarily, push some everyday tasks like coding and debugging back to Claude Opus 4.8.
What's changing
Global rollout, July 1: Fable 5 becomes available again to users everywhere, not just the U.S. organizations that regained partial access to sibling model Mythos 5 on June 26.
New classifiers, narrower target: The updated safeguards are built specifically to catch cybersecurity misuse rather than broadly restricting capability.
Some tasks fall back to Opus 4.8: In the near term, routine coding and debugging requests may be routed to Opus 4.8 rather than Fable 5, as the classifiers are still being tuned.
Ongoing refinement: Anthropic says it will keep adjusting the classifiers over the coming weeks to cut down false positives and more accurately separate genuine misuse from legitimate work.
The bigger picture
Anthropic is also drafting a shared framework with Amazon, Microsoft, Google, and other Project Glasswing partners for rating the severity of AI jailbreaks and coordinating how developers should respond to them — and is inviting other model providers to join.
Separately, the company says it's deepening its collaboration with the U.S. government on model testing, including pre-release access to models and safeguards for evaluation, information-sharing on jailbreaks and misuse, and dedicated joint research resources.
Background
Fable 5 and Mythos 5 — Anthropic's most capable models, sharing the same underlying architecture — were pulled from all users on June 12 after a U.S. export control order required blocking foreign nationals. Anthropic said it lacked a way to verify user nationality in real time and suspended access entirely rather than partially. The directive followed a report from Amazon researchers who found a way to prompt Fable 5 into identifying software vulnerabilities, including a demonstration of how one could be exploited.
Anthropic has said its testing found that several less capable models — including Opus 4.8, GPT-5.5, and Kimi K2.7 — could identify the same vulnerabilities, and that every model tested could reproduce the exploit demonstration. Mythos 5, which offers no such safeguard layer and is aimed at expert cyberdefenders, regained limited access on June 26 for a vetted list of U.S. organizations defending critical infrastructure.
For Pro, Max, Team, and select Enterprise plans, Fable 5 will count toward up to 50% of weekly usage limits through July 7, after which further use will draw on usage credits.
