GitHub is rolling out three coordinated updates that affect how Copilot Business and Enterprise seats are billed, how the Copilot chat experience is delivered, and which code‑review effort level is used by default. All three changes take effect between late September and early October 2026, and they directly impact provisioning pipelines, cost‑tracking, data‑retention policies, and review‑automation settings that engineers and platform teams rely on.
Seat‑based billing for Business and Enterprise
Starting 1 September 2026, new Copilot Business and Enterprise customers can again sign up using credit cards or PayPal. The key operational shift is that a seat must be paid for before the user receives Copilot access. Existing customers using those payment methods will see the same rule applied from 1 October 2026: at the beginning of the next billing cycle each assigned seat incurs an upfront charge. Prices themselves do not change.
Practical effects include:
- Seat assignments that occur mid‑cycle are prorated from the assignment date to the end of the billing period.
- Removing a seat does not generate a refund; the change is reflected in the following month’s invoice.
- Any usage that exceeds the included allowance continues to be purchasable, and spend‑control features, usage tracking, and AI‑credit purchases remain available.
Automation that previously relied on “assign‑first‑use” patterns will need to ensure payment is captured before activation, and cost‑allocation dashboards should be updated to reflect the upfront charge model.
Unified Copilot Chat and cloud‑agent policy
On or after 28 September 2026 GitHub will merge the three existing chat surfaces—Copilot Chat on github.com, the mobile app, and the cloud‑agent—into a single, default‑enabled experience. The separate policies governing each surface are replaced by one unified policy that administrators must review before the launch date.
Key technical points:
- The cloud‑agent will run on a Sandbox environment, promising a faster cloud interaction.
- Chat sessions on github.com will adopt the same agent‑session model, and chat history will be retained for the lifetime of the account rather than the previous 28‑day window.
- Opting out of the unified policy will immediately cut off access to all three chat surfaces after launch.
Teams should audit the new policy settings, verify that any compliance or data‑retention requirements are still met, and adjust any automation that toggles the chat feature on a per‑user basis.
Code‑review effort level default moves to Balanced
GitHub is changing the default effort level for Copilot code‑review from “Lite” to “Balanced” effective 28 September 2026. The default can be set at the organization or repository level; if a repository defines its own level it overrides the organization default. For manually requested reviews, engineers can still pick an effort level from the reviewers bar.
If you prefer to keep the Lite setting, you must explicitly change the default before the rollout date; otherwise the system will automatically apply Balanced. This shift may affect the number of AI‑generated suggestions and the associated token consumption, which in turn could influence usage‑based billing.
Related CloudNinjas coverage: AI engineering.
What This Means For Practitioners
Review your seat‑provisioning scripts to capture payment before activation and to handle prorated charges correctly. Update cost‑monitoring tools to expect upfront seat fees and to track any additional usage purchases. Verify the unified chat policy aligns with your organization’s data‑retention and security guidelines, and decide whether opting out is acceptable for any user groups. Finally, audit organization and repository settings for Copilot code‑review; set the effort level explicitly if the current Lite default is required, and monitor token usage after the Balanced default is applied.
