Anthropic's Claude Fable 5.1 Hits GitHub Copilot but Retains Your Prompts

Anthropic's newest coding model is now live inside GitHub Copilot, carrying a data retention policy that sets it apart from every other Claude model in the lineup. Claude Fable 5.1, part of Anthropic's new Mythos class, is generally available to Copilot Pro+, Max, Business, and Enterprise users, but enabling it means agreeing to Anthropic retaining your prompts and outputs by default so it can run safety classifiers against them.

Built for long-horizon coding work

Claude Fable 5.1 targets long-running, autonomous coding and knowledge-work tasks. GitHub's internal testing showed the model performing well on substantial coding tasks, deep codebase research, and complex agentic feature development. It appears in the Copilot model picker across VS Code, Visual Studio, JetBrains IDEs, Xcode, Eclipse, the Copilot CLI, the coding agent, GitHub Mobile, and github.com. Rollout is gradual, and the policy is off by default, so Copilot Enterprise and Business administrators must enable it in Copilot settings. Billing follows provider list pricing under usage-based billing rather than the flat premium-request model other Copilot models use.

The retention catch

Every other Claude model in Copilot, aside from Fable 5, runs under zero data retention. Fable 5.1 breaks from that: Anthropic retains prompts and outputs to operate safety classifiers that detect harmful use. Retained data is not used for model training, but for regulated industries the distinction between training and retention frequently does not hold up to compliance review.

Anthropic explains the reasoning in its Enterprise Frontier Safeguards announcement: sophisticated misuse often spans many tasks across multiple sessions and accounts, so per-interaction analysis with immediate deletion cannot catch it. Effective detection requires storing data long enough to correlate patterns across time and accounts. The company points to a rise in attempted misuse ranging from fraud to autonomous cyberattacks abusing stolen enterprise credentials as the trigger for introducing 30-day retention starting with Fable 5.

The enterprise escape hatch

Eligible customers can use Claude Fable 5.1 and Claude Fable 5 with zero data retention through the end of the calendar year under a time-bound exemption while Anthropic rolls out Enterprise Frontier Safeguards (EFS). After that, continued zero-retention use requires EFS, which shifts where the data lives rather than removing monitoring.

Under EFS, activity logs used for safety monitoring live in the customer's own Amazon S3, Azure Blob Storage, or Google Cloud Storage bucket, encrypted under the customer's own keys and access policies. Anthropic's automated systems scan a rolling window of that traffic for signals like offensive cyber or biological capability development and leaked credentials, and flags go to the customer's security team rather than an Anthropic reviewer.

Anthropic says it built EFS with more than 100 customers, including the Analysis and Resilience Center for Systemic Risk, whose members include CISOs from Goldman Sachs, Morgan Stanley, Citi, Bank of America, and Wells Fargo. Broad availability is targeted for later this fall, with support across Claude Code, Claude Enterprise, Amazon Bedrock, Google's Agent Platform, and Microsoft Foundry.

Why frontier agents change the math

Frontier coding agents increasingly run unsupervised across large codebases, which is exactly where credential theft and multi-session attacks become difficult to catch with per-request analysis. As agentic capability increases, single-shot stateless safety checks stop being sufficient. The arrangement Anthropic is offering enterprises: retain the privacy story by holding the logs yourself, while a classifier watches the stream.

What to do now

Individual Copilot users get the strongest Claude-family coding option currently available, with the tradeoff that prompts are retained. Administrators managing Copilot for an organization should work through three things before enabling the policy:

1. Determine whether the retention terms clear your compliance bar before enabling the policy. Turning it on constitutes acknowledgment of the retention requirement.
2. If you are in a regulated industry, contact your GitHub account team about zero-retention eligibility. GitHub Support cannot approve this exemption.
3. Plan the EFS transition before year-end if you rely on the temporary exemption, including which cloud account will host the activity logs.

Frontier model vendors are converging on a new default: strong models come with mandatory telemetry, and the negotiation is over who holds the logs rather than whether logs exist.