Claude Fable 5.1 reaches GitHub Copilot with retention caveats

Claude Fable 5.1 is now rolling out in GitHub Copilot for Copilot Pro+, Max, Business, and Enterprise users. The model is positioned for long-running coding and knowledge-work tasks, but its Copilot launch carries a privacy condition that teams need to read before enabling it: prompts and outputs are retained by default for Anthropic’s safety classifiers.

Fable 5.1 joins Copilot’s model picker

GitHub says the rollout is gradual. Individual users can select Fable 5.1 from the model picker where it has arrived; Business and Enterprise administrators must enable the model policy in Copilot settings, and that policy is off by default.

The model is billed at provider list pricing through usage-based billing. That separates it from an included, unlimited plan feature: teams should account for model-request charges when setting budgets and deciding which repositories or workflows can use it.

Anthropic describes Fable 5.1 as its generally available model for ambitious, long-running coding and professional work. The model can operate across large codebases, use tools, recover from failed steps, inspect visual output, and handle diagrams, tables, and PDFs. Those capabilities explain why GitHub is targeting substantial agentic tasks rather than treating the release as another autocomplete option.

The default retention rule changes the deployment decision

When Fable 5.1 is used through Copilot, Anthropic retains prompts and outputs to run safety classifiers. GitHub says the retained data is not used to train Anthropic’s models, but the retention itself still matters for source code, incident data, customer records, and regulated workloads.

Eligible enterprise customers can request zero-data-retention access for Fable 5.1 and Fable 5 through a time-bound exemption that runs through the end of 2026 while Anthropic rolls out Enterprise Frontier Safeguards. Approval is not automatic, and GitHub Support cannot determine eligibility or enable the exception; organizations must work with their GitHub account team and then have an administrator enable the relevant policies.

Anthropic’s model page gives the same underlying constraint in broader terms: Fable 5.1 requires 30-day data retention for safety monitoring by default, while eligible Enterprise Frontier Safeguards customers can use a customer-controlled storage path when available. This is a model-specific exception to the zero-data-retention expectations that apply to most other Claude models in Copilot.

What Copilot teams should check before rollout

Before routing long-running work to Fable 5.1, administrators should:

• confirm that the model appears for the intended Copilot plan and region;
• review usage-based billing and set an explicit budget for model requests;
• decide whether the default retention policy is acceptable for the repositories and prompts involved;
• request the enterprise zero-data-retention path if the organization qualifies; and
• enable the Fable 5.1 policy only after the privacy and cost review is complete.

The practical change is therefore twofold: Copilot gains access to a stronger model for extended autonomous work, while enterprises must treat model selection as a data-governance setting as well as a coding preference. GitHub’s gradual rollout and the year-end retention exemption make the next checkpoint administrative, not merely technical.

Sources and access details

GitHub’s availability announcement lists supported Copilot plans, billing, rollout, administrator controls, and the zero-data-retention exception. Anthropic’s Claude Fable model page documents the model’s capabilities, pricing context, safeguards, and default retention terms.