GitHub has shipped two Copilot governance updates that matter more together than separately.
Enterprise-managed settings can now set a preferred Copilot model as the default for new conversations. At the same time, content exclusions are generally available in the Copilot app and Copilot CLI, so those agentic surfaces respect exclusion policies configured by enterprise, organization, and repository administrators.
The immediate story is administration. The larger story is that Copilot is becoming an enterprise AI operating surface, where model choice and code-context boundaries need to be managed as policy.
Default models move into managed settings
GitHub says enterprises can now choose the default Copilot model that best fits their workflows. The setting applies to new conversations through enterprise-managed settings and is generally available for Copilot Business and Copilot Enterprise in the Copilot app, Copilot CLI, and Visual Studio Code.
The update also supports different defaults by enterprise team. GitHub says administrators can set the model key as overridable and update team configuration files so specific enterprise teams can choose their preferred default model while other users inherit the enterprise default.
That does not replace model-access policy. A default model is a starting point for new conversations. It is still part of a broader control plane that includes which models are permitted, which clients are managed, and which teams are allowed to override central defaults.
Content exclusions now cover app and CLI workflows
The second September 2 update closes a practical gap in agentic coding controls. GitHub says the Copilot app and Copilot CLI now respect content exclusion policies configured by enterprise, organization, and repository administrators.
That matters because command-line and app-based coding agents can touch broad repository context. If excluded files are not honored across those surfaces, a policy that looks strong in an editor can become inconsistent once developers move into more autonomous workflows.
GitHub says the content-exclusion update is available for Copilot Business and Copilot Enterprise customers. The company frames it as a way to help protect sensitive code across agentic workflows.
Copilot governance is getting more concrete
These updates arrive one day after GitHub said Copilot code review can submit pull-request approvals when admins enable the feature. Put together, the direction is clear: Copilot is gaining more agency inside software delivery, and GitHub is adding controls for where that agency is allowed to operate.
For enterprise engineering leaders, the work is becoming less about whether to enable an AI coding assistant at all and more about how to make it predictable. Which model should a team use by default? Which files should never be context? Which repositories can allow AI approval? Which defaults can teams override?
Those questions are mundane, but they are where enterprise AI adoption either scales or stalls.
