On November 18, 2025, GitHub is removing account-level $0 budgets for Copilot premium requests across Enterprise and Team accounts created before August 22, 2025. After this change, overage control shifts to a dedicated policy in your Copilot settings. If that policy is left on its default—Enabled—your organization can rack up paid premium requests beyond the included monthly allowance. The fix is simple but urgent: confirm the policy, set budgets with alerts, and align expectations with engineering leads now.
What exactly changes on November 18, 2025?
Here’s the thing: until now, many enterprises relied on a default $0 budget at the account level to block paid usage when premium requests ran out. On November 18, that automatic $0 budget is removed for eligible enterprise and team accounts. Going forward, premium request overages are controlled by a policy with two states:
• Enabled (default): allow charges when you exceed your allowance.
• Disabled: block premium requests beyond the allowance.
Who’s affected? Enterprise and organization accounts that had a $0 Copilot premium request budget by default. Individual plans (Pro/Pro+) aren’t part of this change. What’s not changing: your included monthly premium request allowance still resets on the first of the month; seat pricing is unchanged; and you still get unlimited code completions and access to standard chat models (subject to rate limits).
Copilot premium requests 101: how allowances actually work
Premium requests are the metered portion of Copilot usage tied to advanced models and features. Each paid seat includes a monthly allowance of premium requests that resets on the first day of the month. Once a user or org burns through that allowance, any additional premium requests become paid usage—unless you explicitly block overages with the policy. You can monitor consumption in real time in the IDE status area and pull detailed reports from your billing settings. Budgets can be set at multiple scopes (enterprise, org, or cost center), and you can also set budgets by SKU (for example, Copilot coding agent or Spark) if you want granular control.
One more point that often gets missed in status reviews: unlimited code completions remain available on paid plans, and you still have chat access without dipping into premium usage for many interactions. The metered part kicks in for specific premium models and agent actions. That’s why policy control matters—so you pay when you intend to, not because of defaults.
Where teams get burned: three high‑risk patterns
In audits this fall, three patterns show up over and over:
1) “Nobody owns the switch.” Finance thinks Engineering has it, and Engineering thinks IT will handle billing policy. No one opens the admin page, and on November 18, the default quietly allows overages.
2) Heavy agent pilots in just a few teams. A small group testing the coding agent can generate a disproportionate share of premium requests, especially if they lean on advanced model chains. Without SKU-level budgets, one pilot can push the whole org over.
3) Multi-org enterprises with uneven governance. Half the orgs disable paid usage; the rest inherit Enabled. Consolidated invoices hide the difference until the bill lands.
Copilot premium requests: the 60‑minute lockdown checklist
If you only have an hour before the change takes effect, this is the path that prevents a surprise bill while keeping your developers productive.
Step 1 — Identify your scope (10 minutes). List the enterprise and organization accounts in scope for Copilot. Confirm who has access to billing and Copilot admin settings. If you’re unsure, schedule a 15-minute huddle with your platform owner or GitHub Enterprise admin.
Step 2 — Decide the policy (10 minutes). Default stance for most enterprises: set premium request paid usage to Disabled at the enterprise level. That blocks paid usage after the allowance. If you already have a clear budget and want to allow overages for specific tools, keep it Enabled but proceed to SKU budgets.
Step 3 — Set budgets and alerts (15 minutes). Create a bundled premium request budget at the enterprise or cost center level with alerts at 75%, 90%, and 100%. If you intend to allow paid usage, set a sensible cap tied to your monthly forecast. If you intend to block paid usage, keep the budget for visibility and alerting.
Step 4 — Create SKU budgets (10 minutes). If you’re piloting the Copilot coding agent or Spark, add dedicated SKU budgets. This isolates pilot costs and prevents one tool from consuming your entire allowance. Start with conservative caps and raise them as you gather data.
Step 5 — Publish the decision (5 minutes). Post a short policy note in your internal handbook or Slack: what’s changing on November 18, what’s allowed, and who to contact for exceptions. Include a link to the usage report dashboard.
Step 6 — Verify from a seat (10 minutes). Ask a developer in the pilot group to confirm they can see the usage indicator and that it reflects the correct policy behavior after your change. Validate that premium requests beyond allowance are either blocked or billable as intended.
How to budget by SKU without wrecking dev flow
Budgeting by SKU gives you control without punishing the wrong users. Start by identifying the tools that actually generate premium requests in your environment—often the coding agent and any heavier model workflows. Assign a modest monthly budget to each SKU aligned to the pilot size. For example, a 25-seat pilot for the coding agent should not be allowed to consume the entire enterprise allowance for 1,000 seats. Lock that with a SKU budget and alerts, while leaving a separate budget for general usage.
Operationally, communicate the guardrails up front: “The coding agent pilot has X allowance this month; if it runs out, we’ll review. Chat and completions remain available.” That message avoids the inevitable panic when someone hits a hard stop. Over time, roll the SKU budgets into your chargeback model or cost center allocations.
Will disabling paid usage break Copilot for my engineers?
No. Disabling paid usage blocks only premium requests beyond the included allowance. Developers still get unlimited code completions and standard chat interactions, subject to rate limits. If your teams rarely hit the premium request ceiling today, they won’t notice a change tomorrow. If a pilot relies heavily on premium features, that team will feel the limit first—which is exactly the point of budget isolation.
Forecasting spend the sensible way
Don’t guess. Pull last month’s usage report, filter to premium requests, and group by SKU and org. Identify outliers—users or teams burning 10x the median—and talk to them. Ask what’s valuable and what’s noise. That conversation tells you whether to raise or tighten budgets. Then set a monthly forecast as: expected usage within allowance + a capped overage for specific SKUs. Revisit weekly for the first month after November 18, then shift to monthly once the pattern stabilizes. If you need help turning this into a repeatable process, our team outlines a pragmatic approach in our guide to avoiding Nov 18 surprise bills.
People also ask
Do we need to change anything if we only use Copilot Chat?
Yes—review your policy regardless. Chat usage is largely outside premium metering, but advanced model calls and agent actions can still consume premium requests. The safest posture is to disable paid usage until you intentionally opt in and set SKU budgets for pilots.
Where do I see who’s burning premium requests?
Open your Copilot billing settings and download the usage report. It shows premium request volume by user, org, and SKU. Cross-reference with your roster of pilots and heavy users. Build a simple dashboard that flags top 10% outliers and tracks spend against budget.
Can I keep a $0 "kill switch"?
You can replicate the effect by setting the premium request paid usage policy to Disabled at the level you care about (enterprise or org). The automatic $0 account budget is going away, but the policy toggle plus budgets achieves the same control with better visibility.
What if we want to allow paid usage, but only for one team?
Keep paid usage Enabled at the enterprise level, then disable it for most orgs and enable it only for the pilot orgs. Use SKU budgets to cap exposure for that pilot. Document the exception so finance knows what to expect on the invoice.
Related changes worth noting
Two nearby updates can affect how you plan rollouts. First, GitHub Actions added support for M2-powered macOS runners and raised limits for nested reusable workflows in early November. If you build iOS or macOS apps or run GPU-accelerated tasks, the performance gains on the new macOS labels are real. Second, the Copilot coding agent no longer requires Actions to be enabled for your org, which gives you flexibility in how you adopt agent features without turning on broader CI/CD functionality. If these touch your stack, review your runner labels and workflow reuse patterns. We’ve covered the Actions angles in our write-up on new Actions limits and M2 runners.
Governance that actually sticks: a lightweight framework
Run this quarterly loop so you don’t have to fire-drill the next policy change:
• Policy: Decide once per quarter if premium request paid usage is Enabled or Disabled by default; document exceptions.
• Budgets: Maintain bundled and SKU budgets with 75/90/100% alerts; adjust caps based on proven value.
• Ownership: Assign a human with authority to Copilot billing and usage review; no orphans.
• Review: Analyze outliers, retire zombie seats, and move pilots to production or sunset them.
• Transparency: Publish a short monthly update—usage vs. allowance, spend vs. budget, action items.
This isn’t heavy governance. It’s hygiene. Fifteen minutes a month costs a lot less than arguing over a surprise line item in Q4.
Edge cases and gotchas
• Multiple enterprises or sandboxes: Policies don’t magically sync. Audit each scope separately.
• Mergers and restructures: When you consolidate orgs, default policies might change. Re-check the toggle and budgets after moves.
• IDE variance: Not every developer sees the same usage indicator behavior. Train teams to check both the IDE indicator and the central report.
• Alerts without action: Alerts at 90% are useless if no one responds. Route them to a shared channel owned by the platform team.
What to do next
1) Today: Open your Copilot admin page and set the premium request paid usage policy to the intended state. If in doubt, Disable.
2) This week: Create bundled and SKU budgets with alerts. Pull usage reports and identify top users by premium requests.
3) This month: Document the policy in your internal handbook, and brief team leads. If you run iOS builds or agent pilots, revisit runner labels and read our note on what to change in Actions now.
4) Ongoing: Review usage monthly, raise or lower budget caps based on clear value, and expand pilots deliberately.
If you’d like help auditing settings or turning this into a repeatable process, start with our what we do page, check our portfolio for similar migrations, or just contact us. We’ve already guided teams through the npm and Actions policy shifts this quarter and can get you to a clean, predictable setup fast.
Zooming out
November is when platform knobs move: npm token lifetimes, Actions limits, runner options, and now Copilot billing defaults. The lesson isn’t "panic every quarter"—it’s to own your settings the way you own your code. Put the premium request policy on a checklist, assign a name to it, and set budgets that reflect a choice, not a default. Do that, and November 18 becomes a non-event.
