On November 18, GitHub is removing legacy $0 account‑level budgets for Copilot premium requests on Enterprise and Team accounts created before August 22, 2025. After that date, whether your org gets blocked or billed comes down to a single policy toggle and the budgets you set. If you haven’t reviewed this, you’re risking either developer outages or surprise invoices. (github.blog)

Here’s the thing: this isn’t a theoretical policy tweak. It will change how your AI usage behaves in the middle of your sprint. Let’s make sure your controls are tight, your comms are clear, and your runway past November 18 is smooth.

Illustration of a billing dashboard with usage bars and a policy toggle

What exactly is changing on November 18?

GitHub will delete the automatically-created $0 premium request budgets for affected enterprise and org accounts. Post-change, your usage beyond the included monthly allowance will be governed by the “Premium request paid usage” policy: Enabled (allow paid overage) or Disabled (block when allowance is exhausted). The change starts November 18, 2025. (github.blog)

If you rely on that $0 budget to block spend today, it won’t save you after November 18. You must either disable paid usage or set explicit budgets and alerts aligned to what your teams actually need. (docs.github.com)

What are Copilot premium requests, and what stays the same?

Premium requests are the metered calls to higher-cost models and capabilities—beyond the included monthly allowance tied to your plan. Allowances reset on the first of each month. Unlimited code completions remain; rate limits may apply, and premium models/features consume the metered pool. The pay‑per‑request path kicks in only if you permit it via policy and budget. (github.blog)

Accounts created before August 22, 2025 historically had a default $0 budget that blocked spend. That specific mechanism is what’s going away; your seats, included allowances, and any non‑zero budgets you’ve already configured remain. (docs.github.com)

Who’s affected—and who isn’t?

Affected: Enterprise and Team orgs that still rely on the legacy account‑level $0 premium request budget. Not affected: Individuals on Pro/Pro+ and any org that already uses the new overage policy with explicit budgets. If you created your account on or after August 22, 2025, you were never on the legacy $0 setup anyway. (github.blog)

Translation: if you’re an enterprise admin who hasn’t touched Copilot billing since summer, assume you’re affected and verify now.

Why this matters to engineering and finance (beyond “billing”)

Premium models and coding agents are increasingly central to code review, refactors, and dev workflow acceleration. A hard block mid‑cycle slows CI, hampers incident mitigation, and tanks morale. The opposite failure mode—unbounded spend—creates a budgeting mess that spikes your blended cost per developer. The fix is governance and right‑sized budgets grounded in actual usage data.

How to fix your Copilot premium requests in 90 minutes

This is the exact checklist we run with clients. Book an hour and a half, grab an engineering manager and a finance partner, and work through it.

1) Pull usage and pinpoint hot spots (15 minutes)

Download premium request usage reports and identify your top 10 consumers, tools driving usage (chat vs. coding agent vs. Spark), and any cost centers. Focus on users who routinely blow past allowance in the first half of the month. (docs.github.com)

Pro tip: Check when spikes occur—release cutoffs, “all hands” refactors, or migration weeks. Your budget shape should reflect those patterns, not a flat monthly average. (docs.github.com)

2) Choose your control model (10 minutes)

You have two main levers: the enterprise policy and the budgets. If you need a hard ceiling, set policy to Disabled to block paid overage. If you want elasticity during crunch time, keep Enabled and set budgets with alerts at 75/90/100%. Bundle across products or use per‑SKU budgets if you want to rein in coding agent separately from chat. (docs.github.com)

3) Set cost centers and budget owners (10 minutes)

Map orgs or business units to cost centers in billing. Assign a budget owner for each who will receive alerts and is empowered to raise caps during releases. This keeps “who approved the overage?” clear and auditable. (docs.github.com)

4) Calibrate the numbers (15 minutes)

Start with last month’s 80th percentile usage for each cost center, then add a 10–20% buffer if your policy allows overage. Turn on “stop usage when budget is reached” for groups where overruns aren’t acceptable (e.g., shared services). Consider separate budgets for the coding agent if your top users lean heavily on it. (docs.github.com)

5) Lock alerts into the workflow (10 minutes)

Route 75/90/100% alerts to a shared channel with finance and the relevant EM. Your goal is zero surprises—someone acknowledges each alert and decides whether to raise the cap or let the block happen. (docs.github.com)

6) Communicate the change (15 minutes)

Send a concise note to developers: what’s changing on November 18, what their IDE will show when limits hit, and who to ping for exceptions. Give them a “request more” link and a reminder that models/features vary in premium request intensity. (docs.github.com)

7) Test a cap before the 18th (15 minutes)

Pick a small group, set a deliberately tight budget, and verify the experience: alert firing, block behavior, and escalation path. Fix gaps now, not during an incident. (docs.github.com)

FAQ: Copilot premium requests, budgets, and policies

Will my developers be blocked on November 18?

If your policy is Disabled or you keep a $0 budget without replacing it, yes—once the included allowance is consumed, premium requests will be blocked. If your policy is Enabled and you have payment set up, usage can continue and incur charges. Check your policy today. (docs.github.com)

Do I need per‑SKU budgets?

Not strictly. A bundled budget is simpler and works for most orgs. Use per‑SKU if you want to constrain a specific tool (for example, an agent that tends to spike during migrations) while keeping chat more open. (docs.github.com)

How often do allowances reset?

Monthly, on the first day of the month. Plan cycles and buffers around that reset. (github.blog)

What if I never changed anything since summer?

You’re probably on the legacy $0 budget if the account predates August 22, 2025. After November 18, that guardrail is gone. Switch to the new policy and set explicit budgets. (github.blog)

Related updates: Actions changes that may affect your plan

This week’s Actions changelog increased nested reusable workflow limits (now up to 10 levels, 50 calls per run) and made M2 macOS runners generally available. If you’ve been splitting pipelines to dodge old limits—or holding off on GPU‑assisted macOS builds—revisit your workflow architecture. These improvements can simplify orchestration and reduce flaky build chains. (github.blog)

Also notable: the Copilot coding agent no longer requires Actions to be enabled for your org. If you disabled Actions strictly to control costs, you can now turn on the agent without reopening CI budgets. Validate policy interactions before you flip the switch. (github.blog)

Team standup reviewing a budget plan on a whiteboard

A simple governance model that actually works

Governance fails when it’s either so strict that teams route around it, or so loose that costs spike quietly. The middle path looks like this:

Set enterprise policy to Enabled for paid usage, but only for cost centers with clear owners and budgets. For shared or experimental groups, keep Disabled. Use bundled budgets by default and per‑SKU budgets only when a tool’s usage profile differs enough to warrant a separate leash. Align alerts to a Slack/MS Teams channel that finance also watches. Review the top 10 users in a monthly ops meeting and adjust buffers before a known heavy period (re-architecture, monorepo migration, quarterly cleanup).

Sample comms you can copy and paste

Subject: Copilot premium requests: policy and budgets before Nov 18

What’s changing: GitHub is removing legacy $0 budgets on Nov 18. Our new policy keeps AI features available while preventing run‑away costs.

What to expect: You have a monthly allowance. If we hit our cost center’s cap, you’ll see a notification in your IDE and requests may be blocked until the budget refreshes or an approver raises the limit.

Need more? Reply here or use the “Request more premium requests” form; include a link to your sprint goal or incident ticket.

Timing: We’ll test this week and finalize settings by EOD Nov 15.

Risk checklist: avoid outages and bill shock

Unknown default? Verify whether your policy is Enabled or Disabled.
No budget owners? Assign one per cost center with raise authority.
Flat budgets? Add buffers for known spikes (releases, migrations).
Silent failures? Turn on 75/90/100% alerts and route to a shared channel.
Unclear messaging? Ship the comms above and pin it in your dev channel.

Data points you can bring to leadership

• Nov 18, 2025: removal of legacy $0 Copilot premium request budgets for Enterprise/Team. Policy now governs overage. (github.blog)

• Monthly allowances reset on the 1st; overage charges apply only if you enable them and have billing configured. (github.blog)

• New Actions limits: up to 10 nested reusable workflows and 50 calls per run; M2 macOS runners now GA. (github.blog)

Practical scenario: the “blocked sprint” you can prevent

Imagine a payments team hitting its allowance on the 20th during a PCI remediation push. With the legacy $0 budget gone, either your policy blocks tools they now depend on, or it quietly allows spend with no alerts. Both are bad. Instead, set a bundled budget sized to last two sprints with 90% alerts, plus a small per‑SKU budget on the coding agent if that’s the spike driver. Give the EM authority to add 15% buffer during incident weeks—with a quick note to finance. (docs.github.com)

What to do next (this week)

Check your Copilot premium request policy and existing budgets. Flip or confirm “Enabled/Disabled” based on your risk tolerance. (docs.github.com)
Download usage reports, identify top consumers, and set initial budgets with alerts. (docs.github.com)
Run a small‑group cap test and validate messaging in IDE and chat. (docs.github.com)
Revisit Actions workflows if you’ve been constrained by nesting limits or waiting on M2 runners. (github.blog)

Zooming out

Every platform is evolving toward metered AI features. GitHub’s shift away from a universal $0 guardrail is part of that trend—more control, but also more responsibility. The orgs that win will treat AI usage like any other utility: forecast it, budget it, and run incident‑style playbooks for spikes.

Do the 90‑minute fix now. Your future self—and your finance partner—will thank you.