More Details on Fable 5's Cyber Safeguards and Our Jailbreak Framework
Anthropic, Thursday, July 2nd, 2026
Anthropic detailed Fable 5's cybersecurity safety classifiers and a draft framework for assessing AI jailbreak severity.
Anthropic detailed the enhanced cybersecurity protections built into the globally redeployed Claude Fable 5.
The company described safety classifiers that detect and block dangerous cybersecurity uses, categorizing activity into prohibited, high-risk dual-use, low-risk dual-use, and benign tiers. Anthropic also introduced an early-draft Cyber Jailbreak Severity framework developed with industry partners to standardize how the severity of AI jailbreaks is assessed and communicated.
The company invited feedback from academia, industry, and government on the framework.