GitHub has announced a set of upcoming changes for Copilot Business and Copilot Enterprise customers, covering seat billing, the GitHub.com and GitHub Mobile experience, cloud-agent access, chat-data retention, and code-review defaults. The changes are scheduled across September and October 2026, so organizations have time to review policy settings, budgets, runner availability, and data-governance requirements.
The most immediate administrative concerns are the scheduled move to a unified Copilot experience no earlier than September 28, the change from Lite to Balanced as the default code-review effort level, and a new billing behavior for covered credit-card and PayPal customers starting October 1.
What changes and when
- September 1, 2026: GitHub plans to begin reenabling sign-ups for new Copilot Business and Copilot Enterprise customers paying by credit card or PayPal.
- No earlier than September 28, 2026: Copilot Chat on GitHub.com, Copilot Chat in GitHub Mobile, and Copilot cloud agent are scheduled to relaunch as one unified Copilot experience.
- September 28, 2026: The default Copilot code-review effort level is scheduled to change from Lite to Balanced for repositories and organizations still using the Default setting.
- October 1, 2026: Existing Business and Enterprise customers using credit card or PayPal are scheduled to incur an upfront charge for all assigned seats at the start of the next billing cycle.
GitHub says Copilot Business and Copilot Enterprise prices are not changing. However, the timing and mechanics of seat charges, included usage, and additional usage may affect budgeting and cash flow.
Billing will require closer seat and usage management
Under the announced change, new Business and Enterprise seat assignments covered by the announcement require payment for each seat before the user receives Copilot access. Existing covered customers will be charged upfront for all assigned seats at the beginning of the next billing cycle starting October 1.
Additional usage may also require payment after included Copilot usage is exceeded. Additional usage remains available for purchase subject to spend controls and usage tracking, and included usage may be prorated during the month to align with seat-cost proration.
Administrators should distinguish this future billing behavior from GitHub’s current general seat-billing documentation. The current Copilot license-change documentation says seats added during a billing cycle are billed on a prorated basis for the remainder of that cycle and provide access immediately. It also says that removing or revoking a seat does not produce a prorated refund, although the user loses access immediately and the reduced seat count is reflected at the applicable billing-cycle boundary.
GitHub’s announcement specifically discusses customers paying by credit card or PayPal. The supplied documentation does not establish how the future behavior applies to contracted, invoiced, usage-based, or other payment arrangements. Organizations using those arrangements should avoid assuming that the October rules apply to them until GitHub documents the scope.
A unified Copilot experience changes policy and retention considerations
No earlier than September 28, GitHub plans to combine Copilot Chat on GitHub.com, Copilot Chat in GitHub Mobile, and Copilot cloud agent into a single Copilot experience with one policy replacing the separate policies. The unified experience is scheduled to be enabled by default. According to the announcement, opting out means users or teams will lose access to Copilot on GitHub.com and GitHub Mobile after launch.
Administrators are instructed to review the policy before September 28 in Copilot settings on GitHub.com. The announced setup path is to select Copilot cloud agent (coming soon) and set the policy option. Current Copilot policy documentation explains that enterprise administrators generally establish policies first, while organizations can make decisions when an enterprise policy permits that choice. Cloud-agent access can also be limited to selected organizations.
The exact post-launch unified policy and all of its administrative controls are not yet documented in the supplied sources. Administrators should therefore review the available setting and verify the resulting behavior after rollout rather than assuming that the current separate-policy model maps exactly to the future experience.
Data retention is another material change. GitHub says that after migration to the agent-sessions experience, GitHub.com chat data will be retained for the life of the account instead of 28 days. Current agent-session documentation says cloud-agent sessions use ephemeral development environments, can run tests and linters, and can be monitored, stopped, or archived. Sessions can be archived but not deleted.
Cloud-agent sessions are shared by default with people who have access to the repository, while local sessions are unshared by default and can be shared with view-only access. Chat and cloud-agent sessions can also exchange context. These behaviors matter for organizations reviewing repository visibility, internal retention rules, and access governance. The supplied sources do not fully document administrator-controlled deletion, export, or access-review options for the retained GitHub.com chat history.
Balanced becomes the code-review default
Copilot code review is scheduled to use the Balanced effort level by default on September 28 for organizations and repositories whose setting remains Default. Administrators who want to preserve Lite must select it explicitly at the organization or repository level before that change.
Balanced routes pull requests to a higher-reasoning model for longer analysis of complex logic, security-sensitive code, and cross-service changes. It uses more AI credits than Lite and may consume marginally more GitHub Actions minutes. GitHub estimates typical AI-credit consumption at $0.05 to $1 for Lite reviews and $0.25 to $5 for Balanced reviews. Those estimates exclude Actions minutes and can change as models evolve. Actual consumption also varies with pull-request size and repository custom instructions.
Organization owners can set a default effort level for automatic reviews, while repository administrators can override the organization default. For a manual review, users can choose the effort level from the Reviewers section of the pull request.
Code review uses AI credits for model interactions and GitHub Actions runners for agentic capabilities such as gathering full-project context and passing suggestions to Copilot cloud agent. Standard GitHub-hosted runners are used by default. Larger hosted runners cost more per minute, while self-hosted runners do not consume GitHub Actions minutes. If GitHub-hosted runners are unavailable or disabled, reviews still run in a more limited form without the additional agentic capabilities; self-hosted runners can be used instead.
Copilot code review is available on paid Copilot plans and is supported on GitHub.com, GitHub CLI, GitHub Mobile, VS Code, Visual Studio, Xcode, JetBrains IDEs, and Azure DevOps in public preview. It is not guaranteed to find every issue or produce correct feedback, so teams should continue to validate suggestions with human review.
What administrators should do
- Review the unified policy before September 28. Check the GitHub.com Copilot settings and confirm that cloud-agent access and organization inheritance match the intended access model.
- Set Lite explicitly if required. Select Lite at the organization or repository level rather than leaving the setting at Default.
- Audit seats and spending. Review assigned seats, billing-cycle timing, included usage, spend controls, and the process for purchasing additional AI usage.
- Model the October billing change. Covered credit-card and PayPal customers should plan for an upfront charge for all assigned seats at the start of the next billing cycle. Do not assume that removing seats will generate a prorated refund.
- Review retention and visibility. Account for life-of-account retention of GitHub.com chat data and the default sharing behavior of cloud-agent sessions.
- Check code-review budgets and runners. Confirm AI-credit paid usage, review policies, Actions runner availability, and who owns the resulting usage. Evaluate Balanced’s higher estimated consumption before making it the default across large repositories.
- Configure access for unlicensed reviewers carefully. Organizations that want members without Copilot licenses to use code review must first enable AI credits paid usage and then enable the policy allowing those members to use Copilot code review on GitHub.com.
Availability and remaining caveats
These are scheduled future changes, not completed rollouts. GitHub has not specified the exact rollout time, rollout scope, or migration mechanics for the unified experience beyond the “no earlier than” September 28 date. The precise implementation of the Sandbox referenced by the announcement is also not fully established by the supplied current sandbox documentation, although GitHub documents organization and enterprise cloud-sandbox access policies and local Copilot CLI controls for filesystem access, networking, and system capabilities.
For teams making compliance or financial plans, the most important open questions concern the complete unified-policy control set, governance options for retained chat data, and billing treatment for payment arrangements outside credit card and PayPal. Those details should be treated as undocumented rather than inferred from the current settings or license-billing pages.
