Claude 5
Top‑performing but controversial model series, forced offline by government order and undermined by a prompt jailbreak; currently unavailable.
Tool overview
Claude 5 shipped as two variants sharing the same base architecture. Fable 5 comes with a safety classifier that downgrades hazardous queries (cybersecurity, biochemical) to older model responses, while Mythos 5 imposes no restrictions and was previously considered too dangerous to release publicly. Within 24 hours of launch, the series achieved a historic lead on the Agent Arena leaderboard, showing dominant performance in task success rates, long‑context reasoning, and complex code generation.
On the third day after launch, hacker collective “Pliny the Liberator” bypassed Fable 5’s security classifier, extracting detailed exploit code and instructions for synthesizing controlled substances. The incident voided Anthropic’s claim of more than 1,000 hours of external red‑teaming with no universal jailbreak found. Shortly afterward, on June 12, 2026, the U.S. government issued an export control directive under national security authority, forcing Anthropic to disable Fable 5 and Mythos 5 for all users worldwide. This was the first known instance of a commercial large model being mandatorily recalled by a government.