Overview
Claude Fable 5 is anthropic’s consumer-facing variant of the Mythos-class cybersecurity model, released June 9, 2026 with additional safety guardrails vs claude-mythos Mythos 5. Shares underlying technology but routes flagged prompts to claude-opus-4-8.
Export Control Saga (June–July 2026)
| Date | Event |
|---|---|
| 2026-06-09 | Fable 5 and Mythos 5 launched |
| 2026-06-12 | US Commerce export controls after Amazon jailbreak report |
| 2026-06-12–30 | Global access suspended (could not verify user nationality) |
| 2026-06-30 | Export controls lifted; new 99% jailbreak classifier deployed |
| 2026-07-01 | Global Fable 5 access restored |
Recent Developments
-
2026-07-24: claude-opus-5 positioned near Fable 5 intelligence at half price; Fable remains stronger on hardest/dual-use tasks per Anthropic (2026-07-24-anthropic-claude-opus-5-launch)
-
2026-07-19/20: Credited by levent-alpoge on a short counterexample to the jacobian-conjecture (jacobian-conjecture-claude-fable-counterexample)
Competitive References
- 2026-07-19: alibaba qwen-38 marketing claims performance “second only to Fable 5” — company self-claim, no independent public bench yet (alibaba-qwen-38-24t-near-frontier)
- 2026-07-16: moonshot-ai kimi-k3 similarly positioned against Fable 5 / GPT-5.6 Sol in vendor/AA narratives
Remediation
- New safety classifier blocks reported jailbreak >99% of trials (Anthropic claim, Amazon researchers endorsed) (2026-07-01-anthropic-fable-mythos-export-ban-therecord)
- project-glasswing industry framework for jailbreak severity scoring (Amazon, Microsoft, Google partners)
- New HackerOne bug bounty program
Related
- claude-mythos
- anthropic
- export-controls
- frontier-ai-governance
- cybersecurity
- amazon
- qwen-38
- kimi-k3
- frontier-models
- jacobian-conjecture
- levent-alpoge
- ai-mathematical-discovery