Claude privacy policy, in plain words
23 min to read the original · checked Nov 16, 2026
Claude uses your chats to train Anthropic's AI unless you opt out, and even then it still uses chats flagged for safety review or ones you rate with a thumbs up or down. It records device and usage details, including your time zone and IP location, and may ask for an ID photo and a face scan to check your age. Anthropic says it does not sell your data.
Sells or trades your dataNo
Uses your data to train AIYes
Tracks you across other appsNo
Collects precise locationYes
Collects biometric dataYes
What they collect
- Name, email and phone number
- Payment details if you subscribe
- Everything you type, upload or connect, and Claude's answers
- Data from apps and services you connect to Claude
- Device, browser, ISP, time zone and IP address, with the location it reveals
- Device and advertising identifiers, including probabilistic ones
- Pages you view, links you click and searches you make
- Error logs, including what was on screen when something broke
- Survey and research interview answers
What they can see
- Every conversation, file and connected-app result you share
- Whole conversations you rate with a thumbs up or down, kept as feedback
- An ID document and a photo or video of your face if you verify your age, with a face geometry template
What they work out about you
- Your location, from your IP address
- Which Anthropic products to recommend to you
Who gets it
- Anthropic affiliates and service providers
- Your employer, if you sign up with a work email linked to its account
- Apps and services you connect, which get your inputs to act on
- Advertising partners for promoting Anthropic, unless you opt out
- Identity checkers such as Persona, in some countries
- Law enforcement and governments when required
- A buyer if Anthropic is merged or sold
What they don't make clear
- "Unless users opt out": your chats train the models by default.
- "Even if you opt-out": chats flagged for safety review are still used for training.
- "We may store the entire related conversation": one thumbs up or down hands over the whole chat.
- "Facial geometry templates": age checks can collect biometric data.
- "We may re-identify": flagged chats used for training can be linked back to you.
- "Publicly available information via the Internet": public web data about you can be in training data.
The main concerns
- Chats train the AI by default
- Flagged and rated chats are used for training even if you opt out
- Age checks can collect an ID and a face scan
- Work-email accounts can be linked to your employer
- Public web data and purchased datasets are used for training
Kept: As long as reasonably necessary; deleted chats are removed from back-end systems within 30 days. Children: Not for under-18s.
Read the full policyWritten from the policy’s own text. How we read policies