Two of the biggest names in tech now offer free AI coding assistants. Google launched Gemini Code Assist for individuals in March 2026 with an astounding 180,000 free completions per month. GitHub, not to be outdone, offers its free tier with 2,000 completions monthly.
Both tools promise to make you a more productive developer. Both also plan to use your code interactions to train their AI models. Here’s exactly what that means for your privacy.
GitHub’s Policy Reversal
On March 25, 2026, GitHub announced changes to its Privacy Statement and Terms of Service. Starting April 24, 2026, interaction data from Copilot Free, Pro, and Pro+ users will be used to train and improve GitHub’s AI models by default.
What counts as “interaction data”? According to GitHub’s policy update:
- Code snippets you submit as prompts
- Outputs and suggestions Copilot generates for you
- Code context surrounding your cursor position
- Comments and documentation in your files
- File names and repository structure
- Your navigation patterns
- Whether you accept, modify, or reject suggestions
- Chat conversations with Copilot
This data may come from private repositories. GitHub is clear: “The interaction data covered by this update (e.g., prompts, suggestions, and code snippets generated during use of Copilot) may be generated while you are working in a private repository.”
GitHub and Microsoft personnel working on AI development will have access to this data. Third-party model providers do not receive it for their own training.
Google’s Fine Print
Google’s approach with Gemini Code Assist for individuals is similarly transparent about data collection:
- Your prompts and related code
- Generated outputs
- Code edits you make
- Feature usage information
- Your feedback on suggestions
Google uses this data “to provide, improve, and develop Google products and services and machine learning technologies.” Human reviewers may read, annotate, and process your data. Google stores disconnected copies (separated from your account) for up to 18 months.
The privacy notice explicitly warns: “don’t submit confidential information or any data you wouldn’t want a reviewer to see or Google to use to improve our products, services, and machine-learning technologies.”
The Numbers: What You Get for Free
The free tier limits differ dramatically:
| Feature | GitHub Copilot Free | Gemini Code Assist Free |
|---|---|---|
| Code completions/month | 2,000 | 180,000 |
| Chat messages | 50/month | 240/day |
| Context window | 128K tokens | 1M tokens |
| SWE-bench score | 33.2% | 63.8% |
| Response speed | Fast | Slow (10+ seconds) |
Gemini offers 90x more completions than Copilot. The 1 million token context window means it can understand larger portions of your codebase. On the SWE-bench benchmark, which tests real-world GitHub issue resolution, Gemini scores nearly double Copilot’s base model.
The tradeoff? Speed. Copilot delivers instant suggestions. Gemini sometimes takes over 10 seconds to respond, which can interrupt your flow.
How to Opt Out
Both services allow you to opt out of AI training data collection while keeping access to the tools.
GitHub Copilot:
- Go to github.com/settings/copilot/features
- Under Privacy, find “Allow GitHub to use my data for AI model training”
- Select Disabled
If you previously opted out, GitHub says your preference has been retained.
Gemini Code Assist: Google’s privacy notice states users can opt out, directing to the “Set up Gemini Code Assist for individuals” documentation. However, the process is less straightforward than GitHub’s single toggle.
Enterprise Tiers Don’t Train
Both companies exempt their paying enterprise customers from this data collection:
- GitHub Copilot Business and Enterprise customers are governed by Data Protection Agreements. Their code is not used for training.
- Gemini Code Assist Standard and Enterprise explicitly state: “Google doesn’t use your prompts or its responses as data to train its models.”
The message is clear: free users pay with their data, enterprise customers pay with money.
What This Means
Both companies are racing to improve their AI coding tools. Training data is the fuel. Free tier users provide that fuel.
This isn’t inherently wrong—it’s a clear value exchange. You get a sophisticated AI coding assistant for free. They get training data to make their models better. Whether that trade works for you depends on what code you’re writing.
If you’re working on open source projects or personal experiments, the privacy implications may be minimal. If you’re writing proprietary code, code that touches sensitive systems, or anything you wouldn’t want human reviewers to see, these free tiers require more careful consideration.
What You Can Do
-
Understand the timing. GitHub’s new policy takes effect April 24, 2026. You have until then to opt out before training begins.
-
Opt out if needed. Both services let you use the AI assistance without contributing training data. The tools still work.
-
Compartmentalize. Some developers use free tiers for open source and personal projects, then switch to enterprise (or no AI) for proprietary work.
-
Consider alternatives. Self-hosted options like Tabby or local models via Ollama keep your code entirely on your machine.
-
Read the policies yourself. Both companies’ privacy documentation is relatively clear. The links are in the sources below.
The “free” in free AI coding assistant has a price. Now you know what it is.