GitHub Copilot will use your code interactions to train AI models from April 24 — here's how to opt out

GitHub has quietly updated its data policies in a way that affects every individual Copilot user. Starting April 24, 2026, your interactions with GitHub Copilot will be used to train and improve GitHub’s AI models by default. That means prompts you send, suggestions you accept or reject, code context, file names, comments, and documentation all become potential training material.

You can opt out. But the default is in.

What exactly is changing

GitHub is updating its Privacy Statement and Terms of Service to formalise what it calls “interaction data” collection for AI model training. This applies to users on Copilot Free, Pro, and Pro+ plans.

The data in scope includes:

• Prompts you send to Copilot
• Suggestions generated and whether you accepted, modified, or dismissed them
• Code context sent alongside your queries
• Comments, documentation, and file names
• Repository structure and navigation patterns

What is not in scope: content from your issues, pull requests, or private repositories sitting passively on GitHub. The key distinction is active use. When Copilot is processing your code to do its job, that interaction is what GitHub wants to learn from.

This data may be shared with GitHub’s affiliates, which includes Microsoft. It will not be shared with third-party AI model providers for their own training purposes, and GitHub is explicit that it does not sell user data.

Who is and isn’t affected

If you use Copilot Business or Copilot Enterprise, nothing changes for you. These plans are governed by a Data Protection Agreement, and that data is not used for model training.

Students and teachers accessing Copilot through GitHub’s education programmes are also exempt.

If you previously opted out of GitHub’s product improvement data collection setting, your preference has been retained. You do not need to do anything.

Everyone else on a personal Copilot plan who has not previously opted out will be opted in automatically from April 24.

How to opt out

It takes about thirty seconds. Go to github.com/settings/copilot/features and look under the Privacy section for the setting labelled “Allow GitHub to use my data for AI model training.” Turn it off.

Opting out has no effect on your access to Copilot or any of its features. You will still get the same suggestions, chat responses, and everything else. You are simply removing your data from the training pipeline.

Why GitHub is doing this

Microsoft’s rationale is straightforward: real-world usage data produces better models. GitHub noted internally that incorporating data from Microsoft employees into the training process led to meaningful improvements in model quality. Extending that to the broader user base is the logical next step from their perspective.

There is precedent for this approach across the AI industry. Most AI products at some point shift from training purely on curated datasets to incorporating signals from actual product usage. GitHub is not doing anything unusual here. The friction comes from the opt-out default rather than opt-in, which puts the responsibility on users to protect their own data rather than asking for explicit consent upfront.

A few things worth knowing if your work involves client code

For freelance developers and small teams on Pro plans, the privacy question extends beyond your own projects. If you are writing code for clients using Copilot, the interaction data flowing through your editor could include snippets from work you do not fully own. That is worth a conversation with clients if confidentiality is part of your working relationship.

Open-source maintainers face a slightly different version of this. Code committed to public repositories already feeds into publicly available training datasets by its nature. Copilot usage traces from working on those same projects could now also enter the pipeline through this policy. It is not necessarily a reason to panic, but it is worth understanding the full picture.

What the community thinks

The reaction in GitHub’s own community discussion thread has been cool. As of publication, the top-level post has collected 59 thumbs-down reactions and three rockets. That is not a scientific measure of developer sentiment, but it is a reasonable signal that this decision has not landed warmly.

There is also a broader irony worth noting. In mid-2025, GitHub open-sourced Copilot Chat. A few months later, it released the source code for Copilot’s inline suggestions. Both were presented as steps toward transparency and developer trust. Defaulting users into AI training data collection shortly after those moves creates a noticeable tension, even if the two things are technically unrelated.

What about EEA and UK users

For users in the European Economic Area and the UK, GitHub has updated its lawful bases section to list AI and machine learning development as a legitimate interest. This is the legal basis it is relying on rather than explicit consent. Legitimate interest processing can still be challenged under GDPR if you believe your data protection rights override GitHub’s interests. The opt-out mechanism above remains the most practical route.

The short version

If you use Copilot on a personal plan, check your privacy settings before April 24. The opt-out is at github.com/settings/copilot/features and takes seconds. If you are on a Business or Enterprise plan, you are not affected. If you previously opted out of related data collection, your preference stands.

GitHub’s goal here is better models, and the logic behind using real interaction data to achieve that is sound. The default-in approach is the part that reasonably bothers people, and the fix is well within your control.