GitHub is changing how Copilot Business and Copilot Enterprise customers manage seats, billing, model access, cloud-agent policy, and code-review defaults.
The changes arrive in stages. Starting September 1, 2026, GitHub says it will begin reenabling new Copilot Business and Enterprise signups paid by credit card or PayPal, with stronger account vetting and updated billing behavior. Starting October 1, existing card or PayPal customers will see upfront charges for assigned seats at the start of the next billing cycle.
No earlier than September 28, GitHub also plans to relaunch Copilot Chat on github.com, Copilot Chat in GitHub Mobile, and GitHub Copilot cloud agent as a single unified Copilot experience and policy.
Seat assignment becomes a billing control
GitHub says all new Copilot Business or Enterprise seat assignments will require payment for each seat before users gain access. If customers exceed included usage, additional payment may be required for continued use.
The prices are not changing, according to GitHub. The practical change is that seat assignment, usage limits, and overage handling become tighter operational controls. Revoking a seat will not create a prorated refund. Adding seats mid-cycle remains prorated.
For administrators, this makes Copilot management look more like other metered enterprise software. Seat inventory, included usage, cost-center ownership, and AI-credit overages need to be monitored together.
Cloud agent and Chat policies are converging
The planned unified Copilot experience folds Copilot cloud agent, Copilot Chat on github.com, and Copilot Chat in GitHub Mobile under one policy. GitHub says the unified experience will be enabled by default after launch.
That policy convergence matters because cloud agents and chat assistants carry different risk profiles. A chat answer may influence work. A cloud agent can operate inside a repository workflow. GitHub says cloud agent will use Sandbox for a faster cloud experience, and Copilot on github.com will migrate to agent sessions, with chat data retained for the life of the account instead of 28 days.
Administrators should review the policy before September 28 if they want to opt out or set a deliberate access posture.
Model policy is becoming more explicit
GitHub’s August 26 global model policy update adds another governance layer. The company says unconfigured and newly generally available models can inherit a default policy state, while explicit enable or disable choices are preserved.
There are important exclusions. GitHub says open-weight models such as DeepSeek and Kimi K2, and models not covered by GitHub’s data retention agreement such as Fable 5, are disabled by default regardless of the global policy.
This is a concrete example of model governance becoming more than a model picker. Data retention, model provenance, default availability, and administrator intent now shape which AI systems employees can use.
