GitHub Copilot just became a three-way AI model marketplace — and the fine print on data retention is the part worth reading
GitHub made Claude Fable 5.1 generally available in Copilot on 1 September and GPT-6 Astra generally available on 4 September, joining Gemini models already in the picker — but Claude Fable 5.1 is the first Claude model in Copilot that retains prompts and outputs by default to run Anthropic's safety classifiers, a data-handling detail buyers should know before picking a model.
12 September 2026
GitHub added two major models to Copilot’s model picker in the space of a week: Claude Fable 5.1 from Anthropic went generally available on 1 September, and GPT-6 Astra from OpenAI followed on 4 September, both rolling out across VS Code, Visual Studio, JetBrains IDEs, Xcode, Eclipse, the Copilot CLI, the coding agent, GitHub Mobile and github.com. Both are pitched at long-horizon, autonomous coding — GPT-6 Astra for planning and independently validating multi-step agentic work, Claude Fable 5.1 for deep codebase research and sustained feature development. Both are available to Copilot Pro+, Max, Business and Enterprise users, billed at provider list pricing under usage-based billing.
The detail that changes the calculus
Buried in GitHub’s own changelog is a data-handling difference worth flagging: Claude Fable 5.1 is the first Claude model in Copilot that retains prompts and outputs by default, specifically to run Anthropic’s safety classifiers that detect harmful use. Anthropic says retained data isn’t used for training, but “retained by default” is a materially different posture from the other Claude models already in Copilot, which don’t require this. For a team with real data-governance obligations — healthcare, finance, anything under a client NDA — that’s not a footnote, it’s a factor in which model gets selected in the picker for which task.
So what
Copilot’s model picker now spans OpenAI, Anthropic and Google models side by side, which is genuinely useful — but “which model is best” and “which model can this team legally use on this codebase” are two different questions, and the second one just got more nuanced. If your organisation has data residency or retention requirements, this is the kind of setting that needs deciding deliberately rather than left on the default the picker loads with. It’s exactly the kind of AI tooling governance we build into project scoping — see our AI-assisted development approach, or get in touch if you want your team’s model choices checked against your compliance requirements before they’re locked in.