GitHub Copilot Will Train on Your Code by Default April 24 — Here's How to Opt Out

GitHub updated its privacy policy on March 25, 2026: starting April 24, GitHub Copilot will train on your interaction data by default. That includes prompts, suggestions you accept or reject, and chat conversations. If you use Copilot — whether on a free plan, individual subscription, or through an employer — you have 26 days to review your settings. Here's exactly what changed, what data is involved, and how to opt out in under two minutes.

What You'll Learn

You'll understand exactly what GitHub is collecting, which plans are affected (hint: it's not just free users), where to find the opt-out settings, and what to watch for after the April 24 deadline.

What Changed

GitHub's March 25 policy update shifts the default from opt-in to opt-out for Copilot training data. Specifically:

• Copilot suggestions feedback (accepted/rejected completions)
• Copilot Chat conversations (your prompts and Copilot's responses)
• Inline chat history in VS Code, JetBrains, and GitHub.com

This applies to: free Copilot plan users, individual Copilot subscribers, and organization members whose admins have not disabled data sharing.

Step 1: Check Your Individual Settings

Go to github.com/settings/copilot.

Scroll to 'Allow GitHub to use my data to improve Copilot.' Toggle it OFF.

Done — this takes effect immediately.

Step 2: If You're on an Organization Plan

Organization admins control data sharing for their team. Go to your organization settings → Copilot → Policies. Under 'Allow GitHub to use organization data to improve Copilot,' set to Disabled.

If you're not the admin, message your IT or DevOps lead and forward this article. The April 24 deadline applies to them too.

Step 3: Verify the Setting Stuck

After toggling off, reload the settings page and confirm the toggle is still off. GitHub has had a history of settings resets with major policy changes — verify after any future Copilot updates.

Step 4: Review What You've Already Sent

If you've been using Copilot Chat, consider what you've shared. Copilot Chat sessions may contain:

• Internal API details, schema names, or function signatures from your codebase
• Business logic descriptions in natural language
• Error messages that reveal stack architecture

This isn't necessarily a data breach — GitHub has privacy safeguards — but it's worth auditing what you've put into Copilot Chat if your codebase is proprietary.

Common Challenges

'I'm on GitHub Free, does this affect me?': Yes. The free Copilot tier is specifically called out in the March 25 update. Free users are opted in by default starting April 24.

'My company manages my GitHub account': Your organization admin controls the data sharing policy for your work account. You may not have an individual override. Escalate to your admin with the GitHub changelog link.

'Does opting out affect Copilot quality?': No. Copilot will continue working exactly as before. You're opting out of contributing your data to future model training — not out of using Copilot itself.

'I already opted out of AI training before — is this different?': The March 2026 update added Copilot Chat specifically. Even if you opted out of code suggestions training previously, check the settings again — Chat data may be a new toggle.

Advanced Tips

For vibe coders using Copilot in CI/CD or automation: GitHub Actions workflows that invoke Copilot APIs may also generate training-eligible data. Check your GitHub Apps settings (github.com/settings/apps/authorizations) for any OAuth apps with Copilot scopes.

Evaluate Claude Code as an alternative: Claude Code does not train on your code by default and operates with a zero-retention API by default on paid plans. For teams with IP sensitivity, this is the practical alternative. See EndOfCoding for a Copilot vs. Claude Code comparison published this week.

The bigger pattern: This is the third major AI tooling data policy change in 2026 (after Cursor's March 2025 zero-retention option and JetBrains AI data clarification in February). Vibe coding teams should have a documented AI data policy — which tools, what data, which opt-outs are in effect. The Vibe Coding Ebook Chapter 19: Security Playbook includes an AI Tool Data Policy template.

For enterprise teams: GitHub Enterprise Cloud has data residency and processing controls that supersede the individual settings. If you're on GHEC, review your MSA data processing addendum — it likely already restricts GitHub from using your data for training.

Conclusion

April 24 is the date GitHub's new default kicks in. The opt-out is genuinely simple — two minutes in settings — but you need to know it exists. Share this with your team. If you're a manager or CTO, this is worth a 10-minute Slack message to your engineers today.

For a comprehensive AI tool data policy template and security checklist for vibe coding teams, see Chapter 19 of the Vibe Coding Ebook. For the latest AI tool comparisons and data policy updates, subscribe at EndOfCoding.