Anthropic opens its most powerful AI to the public with strict guardrails

The rollout marks a significant shift for Anthropic, moving its flagship technology from limited partner testing to broad availability via its API and enterprise plans. Through June 22, the model is included in standard subscriptions, though it will transition to a credit-based model thereafter. To mitigate potential misuse, Anthropic has implemented hard safety limits; in sensitive domains like cybersecurity, biology, and chemistry, the system automatically redirects requests to the older Claude Opus 4.8.

Security remains a primary concern for the company as it eyes public markets. Despite successful stress-testing through external red-teaming and bug bounties, Anthropic is enforcing a 30-day traffic retention policy for all Fable 5 and Mythos 5 users. This requirement, which applies even to firms with previous zero-retention agreements, is framed as a necessary measure to identify novel jailbreak attempts. Pricing is set at $10 per million input tokens and $50 per million output tokens—double the cost of Opus 4.8—a premium that reflects the model's high-level reasoning capabilities and the intensive computational load required for autonomous tasks.