GitHub is introducing three linked changes to Copilot Business and Enterprise: seats must be paid for up front, chat conversations will be retained for the life of the account, and the default code‑review effort level moves from “Lite” to “Balanced.” Each change touches budgeting, compliance, and token consumption, so platform owners need to revisit how they provision seats, audit chat data, and tune review settings.
Upfront Seat Billing and Token‑Based Usage
Effective 1 September 2026, new seat assignments for Copilot Business and Enterprise will require payment before the user can use the service. Existing customers receive a grace period until 1 October 2026. The pricing per seat stays the same, but the billing model now charges at the start of each cycle and adds extra fees if token usage exceeds the included allowance. Because GitHub switched to a token‑driven model in June, the cost of an interaction now depends on the model used and the exact token count, making spreadsheet forecasts for flat per‑seat pricing unreliable. Teams that previously added and removed seats on demand will need to plan seat assignments ahead of the billing cycle and monitor token consumption to avoid surprise charges.
Chat Data Retention Extension
GitHub is consolidating the Copilot Cloud Agent, the web chat UI, and the mobile chat experience into a single interface that will roll out no earlier than 28 September 2026. The policy change that accompanies the rollout extends chat‑message retention from a 28‑day window to the entire lifetime of the GitHub account. This longer retention period directly impacts organizations with data‑retention policies, audit requirements, or contractual obligations about AI‑generated content. Administrators must review the unified policy, confirm the new default, and, if necessary, adjust settings before the change becomes active.
Code Review Default Shift
On the same date, the default review effort level for Copilot‑generated pull‑request feedback changes from “Lite” to “Balanced.” The “Balanced” setting provides more thorough suggestions, which increases the number of tokens consumed per review. Teams that prefer the lighter‑touch approach must explicitly select the “Lite” option before the switch. The deeper default feedback can affect review cycle time and, because token usage is billed, can raise monthly costs.
Related CloudNinjas coverage: AI engineering.
What This Means For Practitioners
Practitioners should take three concrete actions:
- Assign an owner for Copilot seat provisioning; verify that seat assignments are made before the start of each billing cycle and set up monitoring for token usage that exceeds the included quota.
- Audit the new chat‑retention policy; document how long AI‑generated conversations will be stored and ensure that the setting aligns with internal compliance requirements.
- Decide which code‑review effort level matches your team’s workflow; if “Lite” is preferred, update the setting before 28 September 2026 to avoid unintended token consumption.
By addressing these items now, engineering leaders can prevent unexpected invoices, maintain compliance posture, and keep the AI‑assisted development flow predictable.
