Hot AI News

Copilot adds Claude Fable 5.2, but admins must accept data retention first

Claude Fable 5.2 is generally available in GitHub Copilot for four paid plans, but Business and Enterprise admins must switch on a policy that accepts Anthropic's data retention.

Copilot adds Claude Fable 5.2, but admins must accept data retention first
Image: Microsoft

GitHub, Microsoft’s developer platform, has made Claude Fable 5.2 generally available in Copilot. Business and Enterprise administrators must switch on a policy first, which also means accepting that Anthropic keeps prompts and outputs.

Four plans, many editors, and a gradual rollout

According to the changelog, Fable 5.2 is open to Copilot Pro+, Max, Business and Enterprise subscribers. It appears in the model picker of Visual Studio Code, Visual Studio, Copilot CLI, the Copilot coding agent, the Copilot app, github.com, GitHub Mobile on iOS and Android, JetBrains IDEs, Xcode and Eclipse. GitHub says the rollout is gradual, so some users may not see the model yet.

Copilot adds Claude Fable 5.2, but admins must accept data retention first
Image: Microsoft

GitHub describes the model as suited to complex, multistep coding tasks across large codebases. It says early testing showed Fable 5.2 solving more tasks than Claude Fable 5.1 with fewer steps and tool calls, and finishing hard tasks that earlier models missed. The page gives no benchmark figures, so those statements are GitHub’s own and have not been independently verified.

On cost, GitHub says the model is billed at the provider’s list price under usage-based billing and points readers to its Models and pricing page. The changelog itself does not print any rates.

Copilot adds Claude Fable 5.2, but admins must accept data retention first
Image: Microsoft
Advertisements

An off-by-default policy with a retention catch

Copilot Enterprise and Business administrators must enable a Claude Fable 5.2 policy in Copilot settings. It is off by default, and leaving it off keeps the model unavailable to the whole organization. GitHub states that turning it on counts as acknowledging the retention requirement.

That requirement is the unusual part. GitHub says Anthropic retains prompts and outputs to run safety classifiers that detect harmful use, and that the retained data is not used to train Anthropic’s models. Other Claude models, apart from Fable 5, Fable 5.1 and Fable 5.2, still run under zero data retention.

Copilot adds Claude Fable 5.2, but admins must accept data retention first
Image: Microsoft

Some enterprises can avoid retention for now. GitHub says approved customers can keep running all three Fable models without any data being stored, but only until the calendar year ends. The temporary carve-out covers the period while Anthropic builds out Enterprise Frontier Safeguards (EFS). Once it lapses, these models will only be usable with EFS, which adds automated safety monitoring and gives customers more choice over how data is kept, including storage and encryption keys they control.

Eligibility is decided through a GitHub account team, not GitHub Support. Approval does not switch anything on: an administrator must still enable the policy.

Copilot follows Anthropic’s own terms on retention

The retention rule is not a GitHub invention. Anthropic’s own model documentation, as summarized in our earlier report on Fable 5.2 pricing, limits and availability, requires 30-day data retention for the model, as with Fable 5.1, and zero data retention only when Anthropic expressly authorizes it. The same article lists API prices of $8 per million input tokens and $40 per million output tokens, 20% below Fable 5.1’s $10 and $50. Because GitHub bills at list price, those API rates are the closest public reference, though the changelog does not confirm how Copilot converts them.

The same retention conditions appear elsewhere. We reported that Fable 5.2 on AWS GovCloud also comes with data retention strings attached. For Copilot customers, the practical effect is that the safety trade-off now sits in an admin toggle, which security teams in regulated companies will want to review before enabling.

On performance, the only third-party-style number we have covered comes from Cognition, the company behind Devin. Its FrontierCode 2.0 benchmark put Fable 5.2 at 49.1%, ahead of Claude Opus 5.5 at 45.5% and Fable 5.1 at 32.4%, as described in our report on Devin’s benchmark. Cognition sells a coding agent and the result has not been independently reproduced, so it supports GitHub’s claims only loosely.

Advertisements

What the changelog leaves unanswered

Several details are missing. GitHub does not say what Fable 5.2 costs per request inside Copilot, how it counts against plan allowances, or what share of users will see it on day one. It gives no exact date for when EFS becomes mandatory beyond the end of the calendar year, and it says GitHub is working with Anthropic on feedback about EFS without describing any outcome.

Individual Pro+ and Max subscribers face a simpler path, since the policy step applies to Business and Enterprise administrators. The changelog does not spell out whether those individual plans see a retention notice. For organizations, year-end is the date to watch, because the zero-retention exemption lapses then.

Sign Up for the Daily AI Pulse

One email a day. All the stories that matter.

Leo MartinsAI Tools, Prompts & Practical AI Expert

Leo Martins is the AI Tools & Practical AI Expert at Aiholics, focused on helping readers use artificial intelligence to improve productivity, creativity and everyday work.

Leave a Reply

Your email address will not be published. Required fields are marked *

You may also like