Anthropic has introduced a new model classification called "Covered Models," which establishes special data processing rules for the most advanced versions of Claude, including Claude Mythos 5 and Claude Fable 5.
What happened
For Anthropic's most powerful models, such as Claude Mythos 5 and Claude Fable 5, a mandatory 30-day data retention period is being introduced. Under the "Covered Models" policy, the Zero Data Retention (ZDR) option becomes unavailable for these models.
Context
This measure is driven by the increased risks associated with the advanced capabilities of these models in software development, agentic workflows, and cybersecurity. Anthropic categorizes models into "standard" and "high-risk" from a data management and compliance perspective.
Why it matters for the industry
This creates an industry precedent of tiering models based on their functional capabilities and potential risk. It forces developers and companies to rethink their security and compliance architectures, choosing between using top-tier models with mandatory data retention or less powerful versions that support full privacy (ZDR).
Why it matters for users
Users applying the most cutting-edge Claude models to mission-critical tasks will no longer be able to completely disable data retention through standard Enterprise-version settings. Any information processed by models in the "Covered Models" category will be stored in Anthropic's systems for at least 30 days for security review purposes.
Sources
Author
Look at AI, Editorial Team