Will GitHub train AI on my Copilot data?

Yes. Starting April 24, 2026, GitHub will use interaction data from Copilot Free, Pro, and Pro+ users — including inputs, outputs, code snippets, and context — to train AI models unless you manually opt out.

How do I opt out of GitHub Copilot data training?

Go to github.com/settings/copilot/features and disable the AI model training toggle before April 24, 2026. Enterprise and Business plans are excluded by default.

Does GitHub train on my private repository code?

GitHub says it won't use private repository content at rest. However, if you actively use Copilot while working inside a private repo, the prompts, suggestions, and code snippets from those sessions can be collected for training.

Quick Answer

GitHub Copilot Data Training: How to Opt Out (2026)

Published: April 2, 2026

GitHub Copilot Data Training: How to Opt Out

Starting April 24, 2026, GitHub will use your Copilot interaction data to train AI models — unless you opt out. Here’s everything you need to know.

Last verified: April 2026

What’s Changing

On March 25, 2026, GitHub quietly updated its data usage policy:

Before: Copilot Free/Pro/Pro+ users’ interaction data was not used for training
After April 24: All interaction data is used for AI model training by default
What’s collected: Inputs, outputs, code snippets, and associated context from Copilot sessions

Who’s Affected

Plan	Training by Default?	Can Opt Out?
Copilot Free	✅ Yes (after Apr 24)	✅ Yes
Copilot Pro	✅ Yes (after Apr 24)	✅ Yes
Copilot Pro+	✅ Yes (after Apr 24)	✅ Yes
Copilot Business	❌ No	N/A
Copilot Enterprise	❌ No	N/A

Business and Enterprise plans are excluded — their data is not used for training.

How to Opt Out

Go to github.com/settings/copilot/features
Find the “Allow GitHub to use my data for AI model training” toggle
Turn it off
Save your settings

Do this before April 24, 2026 to prevent data collection from starting.

What Data Is Collected

If you don’t opt out, GitHub collects:

Prompts — What you type and the code context sent to Copilot
Suggestions — The code completions and chat responses Copilot generates
Code snippets — The surrounding code context from your editor
Metadata — Language, file type, editor, timestamps

What’s NOT Collected

Private repository code at rest (code just sitting in your repo)
Code you don’t interact with through Copilot
Enterprise/Business plan data

The key distinction: your stored code is safe, but any code that flows through a Copilot session is fair game.

Why This Matters

This is a reversal. GitHub previously promised not to use individual user data for training. The change reflects the broader industry pressure to improve AI models with real-world coding data.

The HN thread about this change had hundreds of comments, with developers expressing concern about proprietary code leaking into models through Copilot sessions — even from private repos.

Alternatives If You’re Concerned

Tool	Data Policy	Price
Claude Code	No training on user data	$20/mo (Pro)
Cursor	Privacy mode available	$20/mo (Pro)
Windsurf	No training on user data	$15/mo
Antigravity IDE	Google’s data policies	Free (preview)

Bottom Line

If you use Copilot Free, Pro, or Pro+, go opt out before April 24. If you’re on Business or Enterprise, you’re unaffected. If this is a dealbreaker, Claude Code and Cursor offer strong alternatives with better data policies.

Last verified: April 2026