GitHub announced on August 28, 2026 that Copilot Chat on github.com, Copilot Chat in GitHub Mobile and the Copilot cloud agent will converge into one experience and policy no earlier than September 28. The same announcement says Copilot code review’s Default setting will move from Lite to Balanced effort on that date.
These are planned changes, not features that are already live everywhere. For QA teams, the important point is that access controls, data retention and automated review behavior may change together even if repositories and test code do not.
What GitHub says will change
- One Copilot policy: separate policies for web chat, mobile chat and cloud agent will be replaced by a unified policy.
- Enabled by default: GitHub says the unified experience will be enabled by default after launch.
- Agent sessions on the web: Copilot on github.com will migrate to the agent-session experience previously used by the cloud agent.
- Longer retention: GitHub says web chat data will be retained for the life of the account instead of 28 days, aligning it with cloud-agent behavior.
- Opt-out consequence: teams that opt out will lose Copilot access on github.com and GitHub Mobile after launch.
- Review default: Default code-review effort will use Balanced instead of Lite. Explicitly selected Lite settings will be preserved.
Why this matters for QA engineers
The GitHub Copilot unified policy changes more than an admin label. A single control can affect where testers start agent sessions, how long their prompts and results remain available, and whether mobile or browser workflows stay accessible.
- Recheck data handling: test evidence can contain customer identifiers, stack traces, URLs and environment details. Teams should confirm that account-lifetime retention fits their policy before using web sessions for triage.
- Validate access paths: test the approved and denied states on github.com and GitHub Mobile so an opt-out does not become an unexpected loss of service.
- Rebaseline automated review: Balanced effort may surface more findings or consume more usage than Lite. Measure useful defects, false positives, review time and cost on representative pull requests.
- Document ownership: identify who can change the organization or repository review-effort default and who approves the unified agent policy.
A short pre-launch QA checklist
- Record the current Copilot policy and code-review effort at organization and repository levels.
- Review whether account-lifetime web-chat retention is acceptable for test artifacts and incident data.
- Run the same set of recent pull requests through Lite and Balanced review; compare actionable findings and noise.
- Test web and mobile access with one permitted account and one restricted account after the policy is available.
- Capture the intended settings and owners before September 28 rather than relying on Default.
GitHub also announced billing changes for credit-card and PayPal customers, including upfront charges for assigned Business and Enterprise seats from the next billing cycle, beginning October 1 for existing customers. Prices are not changing, but QA leads who allocate seats or purchase extra usage should coordinate with their administrators.
Bottom line
GitHub’s August 28 announcement gives QA teams a dated governance and regression-testing task. Before the unified Copilot experience launches, review retention, verify access behavior and explicitly choose the code-review effort that matches your quality and usage targets.