Mythos 5 safeguards block specific high-risk areas in the Fable variant — first frontier-lab disclosure that public-tier safety differs from private-tier capability
Anthropic's disclosure that Claude Fable 5 ships with safeguards that block responses in specific high-risk areas — enabling public distribution — formalizes a two-tier safety regime. Mythos 5 to approved Glasswing partners retains the full capability surface; Fable 5 to public subscribers operates inside a narrower behavioral envelope. The two-tier model is the first time a frontier lab has explicitly disclosed differential safety conditioning by customer tier.
The two-tier safety pattern is the substantive shift. Frontier-lab safety historically operated as a binary — model ships with safety conditioning or it doesn't. Mythos 5 / Fable 5 introduces conditioning that varies by customer authorization: Glasswing partners (named enterprises with explicit access agreements and disclosure obligations) get the wider envelope; public subscribers (Pro/Max/Team/Enterprise self-serve) get the narrower one. Both are the same underlying weights with different deployment-time safety layers.
The disclosure obligation is the regulatory frame. EU AI Act Article 5 prohibitions and the August 2 transparency obligations apply to the deployed system — meaning Fable 5 (public) and Mythos 5 (Glasswing) are arguably two different systems for compliance purposes, even though they share a model artifact. That's a clarifying signal for the GPAI Code-of-Practice debate: provenance is tied to the deployed product surface, not just the underlying weights. The catalog launch coverage shows the product-side structure; the regulatory implications are downstream.
Anthropic — Claude Fable 5 and Claude Mythos 5 → · SD Times — Anthropic releases Claude Fable 5, Mythos 5 → · CodersEra — Claude Fable 5: Anthropic's New Mythos-Class Model (Benchmarks, Pricing & What's New) →