Skip to content

See the new AI dev workflow first at NTL DEPLOY. Join live October 1st →

Security and privacy for AI features

Learn how we process data, work with AI models, and handle data storage for Netlify’s AI features.

Netlify’s AI features include the following:

Note that Ask Netlify AI is enabled by default for all site visitors to docs.netlify.com and is available in-app for all users of Netlify.

Agent Runners, AI Gateway, and AI-assisted troubleshooting for failed deploys can be disabled for all projects in your Netlify team account from your Netlify dashboard.

For Enterprise plans, these features are turned off by default. For other plans, these features can be turned off at any time.

To disable Netlify’s AI features, check out Manage AI features.

AI primitives, such as AI Gateway, are only activated if you explicitly add this capability in your project’s code.

Netlify’s AI features exclusively use third-party providers for inference. These providers are not permitted to train their models on inputs received through Netlify.

Different features handle storage differently, based on their purpose:

  • “Why Did It Fail?”
    • Uses build logs already stored in our system. When invoked (only for failed deploys), a sample of logs is sent to an AI provider, and the resulting suggestion is stored for the customer for up to 30 days.
  • Agent Runners
    • Provides a controlled environment to run AI agents (e.g., Claude Code, OpenAI Codex, Gemini CLI). Logs from these sessions are captured for display to the user, troubleshooting, and audit purposes, under the same policies that cover build logs and data for our CI system.
  • AI Gateway
    • Acts as a proxy for routing inference requests to providers. It does not store inputs or outputs, except for:
      • Metadata about requests (which does not include prompts or generated output) used for logging, auditing, and support.
      • Enhanced logging is available - upon customer request only - for troubleshooting.
  • Ask Netlify
    • Uses an external provider (Kapa.ai) to answer questions about Netlify and web development. Questions and answers are stored to help improve support and product features.

We run evaluation tests on both the models and our system inputs to reduce the risk of unsafe results for our AI features.

However, customers are responsible for verifying the outputs of models used in their applications.

Our AI capabilities are powered by trusted providers, including:

  • OpenAI
  • Anthropic
  • Google
  • Kapa.ai (used exclusively for the Ask Netlify AI feature in the Netlify dashboard and on the Netlify docs site.)