On September 16, 2026, GitHub made Copilot budget increase requests generally available. Previously, a member who used all available Copilot AI credits was blocked from features that consume credits. The new flow lets that member request more budget at the limit, while an authorized owner or billing manager can approve, adjust, or deny the request in settings.
Requests automatically route to the account that pays for the budget. An organization-owned budget appears in organization settings, while an enterprise-owned budget appears in enterprise settings. That connects the user’s need with the account responsible for payment and gives the budget owner visibility into where additional demand is coming from.
An organization owner, enterprise owner, or billing manager can review pending requests under Requests from members, set a new amount, and select Approve and increase. Approval updates the member’s budget and immediately restores access to AI credits. The ability to adjust or deny means that reaching the limit does not become unlimited access by default.
The feature is available for GitHub Copilot Business and GitHub Copilot Enterprise under usage-based billing. It is not a model update, and GitHub does not claim that it reduces total spend. It addresses an operational tradeoff: how to avoid interrupting a coding agent or AI tool while keeping additional usage subject to a deliberate spending decision.
For Copilot teams, a budget request is part of agent governance. When AI can keep analyzing, editing, and testing, credits become both a workflow constraint and a cost signal. Making an increase a recorded approval event is easier to audit than simply raising a global ceiling, and it creates a way to compare why teams need more capacity with what they delivered.
Before rollout, teams can define a budget owner, response-time target, per-team allowance, unusual-usage alert, and reason for approval. They can also review credit consumption alongside accepted code changes, test outcomes, or delivery time. These are operating recommendations, not performance data from GitHub; the actual policy should match the organization’s billing, access, and risk requirements.
Permissions still need separation. Someone who can approve a budget increase should not automatically receive repository write access, deployment access, or production access. Restoring credits only restores access to Copilot usage; it does not grant the model more data or more actions. Separating billing, code permissions, and deployment approvals limits the power concentrated in one account.
The update reflects a more operational phase of enterprise AI. Cost control is no longer only a month-end invoice exercise; it becomes a decision during an agent run. GitHub provides the request, approval, and access-restoration workflow. Organizations still need to measure the value, error rate, and risk of each increase before deciding which AI-assisted work deserves to scale.
