Loading prices…
🩸BEARISH

Anthropic launches Fable 5: first Mythos-class AI model

The Mythos tier is the headline, but the auto-routing layer that funnels cyberattack and code-extraction queries down to Opus 4.8 is what DeFi teams will be reading closely.

Anthropic has launched Claude Fable 5, its first publicly available Mythos-class AI model. To manage the risk profile of a model at that capability tier, Fable 5 ships with an automatic safety-routing layer: any query that touches cyberattacks, biological or chemical threats, or model distillation is shunted down to Claude Opus 4.8, the same model credited with surfacing a critical vulnerability in Zcash's Orchard protocol.

Why it matters

The Mythos designation is Anthropic's signal tier for models with elevated capabilities — and the Fable line is the first one the company has cleared for general release rather than restricted access. The safety router is the operative feature: it is what determines whether a Mythos-tier model can be pointed at a smart contract, an exploit chain, or a privacy-coin wallet and get a useful answer back. Officially, the answer is no — those queries fall through to Opus 4.8, which the launch blog notes previously surfaced the Zcash Orchard bug.

That distinction is awkward: the same model class that found a real critical in a privacy coin is now the fallback for any query touching offensive security. For DeFi teams, the practical read is that the ceiling for AI-assisted auditing just got a named model, while the ceiling for AI-assisted exploitation sits one tier below — and the gap between the two is the part defenders will be auditing.

Market impact

The launch lands amid a broader push to use frontier models for smart-contract auditing, where defenders hope AI can triage bug bounties and formal-verification gaps faster than human teams alone. Fable 5 lifts that ceiling; the routing policy sets the floor for offensive use.

Source: [Claude Fable 5 and Claude Mythos 5](https://www.anthropic.com/news/claude-fable-5-mythos-5)

Frequently asked questions

  1. What is Claude Fable 5?

    Fable 5 is Anthropic's first publicly available Mythos-class model — the company's elevated-capability tier — released for general use with an automatic safety-routing layer built in.

  2. What does the Fable 5 safety router do?

    It redirects any query touching cyberattacks, biological or chemical threats, or model distillation to Claude Opus 4.8, a less capable model that previously surfaced a critical vulnerability in Zcash's Orchard protocol.

  3. Why is the Fable 5 launch relevant to DeFi security?

    The release sets a new ceiling for AI-assisted smart-contract auditing while pinning offensive cyberattack queries to a lower tier, giving defenders and attackers a clearly drawn capability boundary.

  4. Did Fable 5 find a real Zcash vulnerability?

    No — Opus 4.8 did. The launch material credits Opus 4.8 with surfacing a critical vulnerability in Zcash's Orchard protocol, and Opus 4.8 is the model that cyberattack-flavored queries on Fable 5 are routed to.

  5. What should DeFi teams watch after the Fable 5 launch?

    Whether other frontier labs adopt comparable routing policies, and whether any DeFi protocol lists Fable 5 in its audit pipeline — both would signal how the Mythos tier gets absorbed into the security stack.

Source attribution
Aggregated from WuBlockchain · Verified · Last refreshed 45d ago
Open original →