In a Nutshell: Anthropic splits Claude Fable 5 into a public version (with safeguards) and a restrictive version (Claude Mythos 5 without security layers) for verified cybersecurity experts.
On June 9, Anthropic made Claude Fable 5 generally available, implementing a two-tier model concept: the public version runs with security classifiers, while Claude Mythos 5 – the same base model without these safeguards – remains reserved for a verified group of cybersecurity professionals.
According to Anthropic, Claude Fable 5 is the most capable model the company has developed to date. Rather than training two entirely different versions, Anthropic implements a hybrid approach: a single base model is differentiated through different security classifiers to create two distinct products.
The public variant (Claude Fable 5) is shipped with cybersecurity safeguards enabled. With this, Anthropic aims to limit potential for abuse – such as malware development or exploits – while simultaneously making the model’s capabilities available to the broad user base.
Claude Mythos 5, the counterpart without these security layers, remains accessible to a verified circle of cybersecurity professionals. This approach enables qualified security researchers and penetration testers to leverage the full model capacity for defensive purposes – such as vulnerability analysis or testing attack scenarios – without these capabilities being available in the general public version.
Source: thehackernews.com · Published June 10, 2026
Lumi AI News — AI-assisted curation in accordance with Article 50 EU AI Act. Paraphrase and classification by Lumi News Pipeline v1.6.5.