Anthropic’s latest policy changes mean that chats and coding sessions with Claude—unless you intervene—are now eligible for use in future AI model training. This shift applies to users of Claude Free, Pro, and Max plans, and it introduces a five-year data retention period for those who permit data use. If you want to stop your conversations from being used in model training, you need to actively opt out. Here’s how to do it, along with important context on what these changes mean for your privacy.

Opt Out of Claude AI Model Training (Most Effective Method)

Step 1: Sign in to your Claude account on the web or mobile app. After the policy update, existing users will see a pop-up titled "Updates to Consumer Terms and Policies." This pop-up explains the new rules and includes a toggle labeled "Help improve Claude" or similar. By default, this toggle is set to allow data use for training.

Step 2: Carefully review the pop-up before clicking "Accept." Make sure the toggle for model improvement is switched off. The toggle should show an X or be set to the left, indicating opt-out status. Only after confirming the toggle is off should you proceed to accept the terms. This action ensures your new and resumed chats are excluded from training.

Step 3: If you’ve already accepted the new terms or missed the toggle, you can still change your preference. Open Claude, then navigate to Settings > Privacy > Privacy Settings. Find the "Help improve Claude" or similar toggle and set it to off. This setting change will prevent future chats and coding sessions from being used in model training. However, any data collected while the setting was on may already be included in ongoing or completed model training cycles.

Step 4: For added privacy, regularly delete individual chats you don’t want retained. Deleting a conversation ensures it won’t be used in future training runs. Remember, deleted chats are excluded from new model training, but may remain in models already trained on them.

Step 5: Review your settings periodically, especially after major updates or policy changes. Companies may update default settings or introduce new options. It’s important to confirm your preferences remain as intended.

Use Enterprise or Commercial Accounts for Stronger Data Protections

Anthropic’s data training policies do not apply to Claude for Work, Claude Gov, Claude for Education, or API access (including via Amazon Bedrock or Google Cloud Vertex AI). These plans operate under separate commercial terms that prevent your chats from being used for model training. If you require a higher level of data privacy—such as for business or regulated environments—consider upgrading to a commercial or enterprise account. This option is more costly but offers contractual data protection not available on consumer plans.

Run AI Models Locally for Maximum Privacy

If you want to ensure your data never leaves your device, running AI models locally is the most robust approach. This method requires purchasing suitable hardware (such as high-performance GPUs), downloading open-source models, and configuring them on your machine. While this option provides total control over your data, it comes with trade-offs: higher upfront costs, technical setup, and potential limitations in model capabilities compared to cloud-based services. For those with strict privacy needs or technical expertise, local deployment is the only way to guarantee no third-party data collection or retention.

Key Considerations and Cautions

Anthropic’s opt-out system defaults to data sharing, so users who quickly accept new terms may inadvertently allow their chats to be used for training. Always check for small toggles or settings during account updates or signups. Note that opting out only affects new and resumed chats; data from sessions where sharing was enabled may already be included in existing models and cannot be retroactively removed.

Be aware that data retention for opted-in users is five years, while those who opt out revert to a 30-day retention window. Deleting specific chats or your account will prevent those conversations from being used in future model training, but may not remove them from models already trained on that data.

For users highly concerned about privacy, document your opt-out actions (such as taking screenshots of settings) in case you need to prove your choices in the event of a future dispute or class action.

By proactively managing your privacy settings or choosing alternative solutions, you can stop your Claude AI chats from being used to train Anthropic’s models and keep greater control over your personal data.