AI Privacy Blog

Your Code Is Now AI Training Data: GitHub Copilot's Quiet Opt-Out Switch

April 8, 2026 · 5 min read

Starting April 24, GitHub will use your Copilot interactions to train AI models by default — including code snippets from private repos. If you don't explicitly opt out, your code becomes training data. Here's what changed, why it matters, and how to protect yourself.

By the GPTAnon Editorial Team | April 8, 2026

---

If you use GitHub Copilot on a Free, Pro, or Pro+ plan, here's something you need to know right now: starting April 24, 2026, GitHub will use your Copilot interaction data — your prompts, code snippets, suggestions, and surrounding context — to train its AI models. By default. Unless you opt out.

That's not a drill. That's a policy update buried in a Terms of Service revision that most developers will never read.

What Changed — and Why It Matters

Previously, GitHub Copilot did not use individual users' interaction data for model training unless you explicitly opted in. That was the deal. You got AI-assisted coding, and your code stayed yours.

Now, GitHub has flipped the switch. The default is opt-out, not opt-in. If you do nothing by April 24, your coding sessions become fair game for training future AI models.

Let's be clear about what "interaction data" means here. It's not just the suggestion Copilot gives you. It includes:

Code snippets you write and Copilot processes

Prompts you send to Copilot Chat
Surrounding context — the code around your cursor, file names, repo structure
Your feedback — thumbs up/down on suggestions
Navigation patterns and comments in your files

That's a lot of signal about how you write code, what you're building, and how your projects are structured.

"But They Said They Don't Train on Private Repos"

Here's where it gets slippery. GitHub is careful to say they don't train on private repository content at rest. That sounds reassuring until you parse it.

When you're actively using Copilot in a private repo — writing code, accepting suggestions, chatting with Copilot — all of that interaction data is generated from your private codebase. GitHub processes your private code to run Copilot, and that processing creates the interaction data they now want to use for training.

So no, they're not crawling your private repos at rest. But the moment you use Copilot, your private code flows through a pipeline that now feeds into model training. The distinction is more legal than practical.

> If your code tools can't respect your privacy, your AI tools should. Try GPTAnon — anonymous AI chat with no data collection →

As The Register put it, GitHub is essentially training on your data after all — just with extra steps and careful wording.

The Developer Backlash Is Real

The Hacker News thread on this change exploded with hundreds of comments. Developers working on proprietary software are rightfully asking: does this mean my company's IP is flowing into a training pipeline?

For solo developers and open-source contributors, the stakes are more personal — it's about consent and control. The entire point of an opt-in model is that you choose to participate. Flipping to opt-out means GitHub is banking on inertia. They know most people won't change their settings. That's not consent. That's a dark pattern.

Who's Affected (and Who Isn't)

Affected: All Copilot Free, Pro, and Pro+ individual users.

Not affected: Copilot Business and Enterprise users — their data was already excluded from training. Also, if your GitHub account belongs to a paid organization, your interaction data is excluded even if you use a personal Copilot plan.

How to Opt Out (Do This Now)

If you want to keep your code out of GitHub's training pipeline:

Go to github.com/settings/copilot

Under the Privacy section, find "Allow GitHub to use my data for AI model training"

Set it to Disabled

Do this before April 24, 2026

It takes 30 seconds. Don't wait.

The Bigger Picture

This move by GitHub is part of a broader trend we've been tracking at GPTAnon: AI companies gradually expanding their data collection defaults, betting that users won't notice or won't bother to change settings.

We saw it with Meta's AI training on Instagram posts. We saw it with LinkedIn quietly training on user content. Now we're seeing it with the very code developers write to build software.

The pattern is always the same: launch with privacy-friendly defaults, build trust and market share, then quietly shift to opt-out once you have a captive user base. It's the exact problem GPTAnon was designed to solve — giving you access to powerful AI models without the bait-and-switch on your data.

If you care about code privacy, the time to act is now. Opt out, spread the word, and demand better from the tools you depend on.

---

Sources:

---

Tired of AI companies quietly harvesting your data? GPTAnon lets you use 25+ AI models — GPT-5, Claude, Gemini, DeepSeek, and more — without creating an account, without logging your conversations, and without ever training on your inputs. Start a private conversation →