Mythos Returns: Anthropic’s Tests Show Fable 5 Wasn’t Uniquely Risky
Anthropic says inside testing discovered Claude Fable 5 posed no distinctive cybersecurity hazard, as Claude Mythos 5 returns globally on July 2.
The admission accompanies Fable 5’s world relaunch, capping an 18-day suspension triggered by US export controls on June 12. Anthropic examined rival fashions to gauge the true risk behind the restriction.
Why Anthropic Suspended Fable 5
Fable 5 and Mythos 5 launched June 9, sharing the identical core mannequin with the previous open to the general public. Mythos 5 stayed restricted to a small variety of trusted Project Glasswing companions for defensive cybersecurity work.
The export controls arrived after Amazon researchers discovered a method to bypass Fable 5’s safeguards. The method prompted the mannequin to determine software program vulnerabilities and, in a single case, show an exploit.
Anthropic’s tests discovered that Claude Opus 4.8, GPT-5.5, and Kimi K2.7 may determine the identical vulnerabilities Fable 5 flagged within the Amazon report. Every mannequin examined may reproduce the one exploit demonstration too.
The discovering suggests the directive focused a spot shared throughout the trade, not a Fable-specific risk. Anthropic nonetheless constructed a stronger classifier to dam the method, which now additionally flags extra routine coding and debugging requests.
How The Guardrails Actually Work
Fable 5 launched with the strongest security margin Anthropic has constructed into any mannequin. Its classifiers block requests that look even barely dangerous, not simply the clearly dangerous ones. The new classifier educated after the Amazon report blocks the reported bypass in over 99% of instances, (*5*). Blocked requests now reroute robotically to Opus 4.8.
That security margin comes at a price. Anthropic acknowledges the classifier flags extra benign coding and debugging requests, and says it’ll maintain tuning it to chop false positives. Mythos 5, which carries fewer of those guardrails, returned just for Mythos 5 institutions cleared by the federal government on June 26.
Anthropic’s personal knowledge raises a tougher query. If weaker fashions can already do what Fable 5 was banned for, what normal will regulators apply the subsequent time a frontier mannequin launches?
The publish Mythos Returns: Anthropic’s Tests Show Fable 5 Wasn’t Uniquely Risky appeared first on BeInCrypto.
