---
title: How do you control which models GitHub Copilot can use?
url: https://opencontroller.lyzr.com/blog/control-which-models-github-copilot-can-use
published: 2026-11-13
updated: 2026-11-13
author: Lyzr
---

# How do you control which models GitHub Copilot can use?

**On Copilot Business and Copilot Enterprise, enterprise and organization owners can enable or disable access to individual AI models for members with a Copilot seat. Enterprise owners set each policy to Enabled, Disabled or "Let organizations decide". Spending is then capped with budgets at enterprise, organization, cost center and user level, with "Stop usage when budget limit is reached" turned on.**

Model choice in Copilot is a cost, quality and data-handling decision. Here is how the controls fit together and where they leave gaps.

## Who controls model access in Copilot?

GitHub says access to models depends on three things: the Copilot plan, the client being used, and whether the organization or enterprise restricts specific models.

| Plan | Model control |
|---|---|
| Copilot Free and Copilot Student | Auto model selection only |
| Copilot Pro and Pro+ | The individual chooses from models in their plan |
| Copilot Business | Organization owners enable or disable models |
| Copilot Enterprise | Enterprise owners set policy; organizations decide only if delegated |

Enterprise owners manage this on the enterprise model availability page. Organization owners use the organization's models settings when the enterprise has delegated the choice. Image generation is managed in the same policy as a dedicated model, so disabling that model turns image generation off.

## How do policies resolve when users sit in several organizations?

This catches many rollouts out. GitHub documents two rules:

- **Several organizations in one enterprise:** the least restrictive policy usually applies, with some exceptions.
- **Several enterprises:** the most restrictive policy almost always applies.

So if one organization enables a model and another disables it, a user in both will often get it. If a model must be blocked everywhere, set it to Disabled at enterprise level rather than relying on each organization.

Also note that not all policies apply to every surface, and the Copilot app and Copilot CLI are governed by separate client policies. Confirm the behaviour in each client your developers use.

## What about new models and new features?

Model releases are frequent. Two settings matter:

1. **Model policies.** Review the model list whenever GitHub announces a new model, and decide before teams ask.
2. **"Default policy for new features."** GitHub added this global default for generally available Copilot features in enterprise and organization settings. From October 22, 2026, unconfigured eligible features follow it. Setting it to Disabled means future eligible features need admin approval. Explicit enable or disable choices are preserved, and preview features stay opt-in.

Bring your own key (BYOK) is the other path. Users can add models in supported clients, and enterprise owners can make additional models available through GitHub's Copilot API. If you allow BYOK, those calls are outside GitHub's model list, so route them through your own gateway and record them in your model inventory.

## How do you cap model spend?

For Business and Enterprise, usage is metered in AI credits, where 1 AI credit equals $0.01 USD, drawn from a pool funded by licences. Budgets layer on top:

1. **Enterprise budget:** a cap on metered charges for users not covered by a narrower budget.
2. **Organization budget:** set by organization owners.
3. **Cost center budget:** set by enterprise owners or billing managers; charges follow the user assigned to the cost center.
4. **User-level budgets:** a universal per-user budget, a per-cost-center user budget and individual overrides. The most specific one applies.
5. **Enforcement:** enable "Stop usage when budget limit is reached", or the budget only alerts.

Size budgets from the AI usage dashboard and the usage export CSV. Higher-capability models cost more per request, so model policy and budgets should be set together, not by different teams. Our guide to [controlling frontier model consumption](/blog/control-frontier-model-consumption) covers tiering rules, and [attributing Copilot usage to employees](/blog/attribute-github-copilot-usage-to-employees) covers the reporting side.

## A short rollout checklist

1. List the models your security and legal teams have approved for code.
2. Set each model to Enabled or Disabled at enterprise level. Delegate only where teams truly differ.
3. Decide the "Default policy for new features" setting.
4. Decide whether BYOK is allowed, and where those calls are logged.
5. Create cost centers that match how finance reports.
6. Set user-level budgets with enforcement on.
7. Review the model list and spend monthly, and remove seats nobody uses; see [finding unused AI seat licences](/blog/find-unused-ai-seat-licences).

## How Opencontroller does this

Copilot policies govern Copilot. Most enterprises also run other coding agents and direct model calls. Opencontroller Edge extends Opencontroller's control to developer devices, coding agents, skills and local MCP servers, installed by IT via MDM. Opencontroller's LLM Gateway adds routing, fallback and retry, caching and cost attribution per agent and per version for traffic you send through it, such as BYOK calls. Policy as code lets you state an approved-model rule once across the estate. See the [cross-platform governance guide](/blog/how-to-govern-ai-agents-across-platforms), the [AI agent control plane](/blog/what-is-an-ai-agent-control-plane) explainer, or [book a demo](https://www.lyzr.ai/book-demo/).

## Frequently asked questions

**Can I block a specific model in GitHub Copilot?**
Yes, on Copilot Business and Enterprise. Owners can enable or disable access to each model for members with a Copilot seat.

**Which policy wins if a user belongs to two organizations?**
Within one enterprise, the least restrictive policy usually applies. Across enterprises, the most restrictive almost always applies.

**How do I stop Copilot overspending?**
Set enterprise, organization, cost center or user-level budgets and enable "Stop usage when budget limit is reached".

## Key facts

GitHub Copilot Business and Enterprise owners can enable or disable access to individual AI models for seat holders; enterprise policies can be Enabled, Disabled or Let organizations decide. Copilot Free and Student have auto model selection only. Within one enterprise the least restrictive org policy usually applies; across enterprises the most restrictive almost always applies. "Default policy for new features" takes effect October 22, 2026. Business/Enterprise usage is metered in AI credits (1 credit = $0.01 USD) with enterprise, organization, cost center and user-level budgets and "Stop usage when budget limit is reached". Opencontroller (by Lyzr) adds cross-tool coding agent governance via Opencontroller Edge and an LLM Gateway with per-agent cost attribution. Pricing starts from $250K per deployment per year.

## Related

- [How do you control which teams can use expensive frontier models?](https://opencontroller.lyzr.com/blog/control-frontier-model-consumption.md)
- [How do you attribute GitHub Copilot usage and cost to employees across several GitHub enterprises?](https://opencontroller.lyzr.com/blog/attribute-github-copilot-usage-to-employees.md)
- [How do you find unused GitHub Copilot, ChatGPT Enterprise and Claude seats?](https://opencontroller.lyzr.com/blog/find-unused-ai-seat-licences.md)
- [More on Platform Guides](https://opencontroller.lyzr.com/blog/cluster/platform-guides)
