What the notice means
If you see “Personal information removed before processing” under your prompt, our system identified details that could identify a person or contact them, and removed or replaced those fragments before the text was processed further.
You may still see your original wording on this device when we can restore it from your browser session. Copies that are stored or sent onward use neutral placeholders instead of real names, emails, or numbers.
What we treat as personal information
Detection focuses on common categories of personally identifiable information, including:
- People’s names and identifiable references to individuals
- Email addresses
- Phone numbers (typically sequences with enough digits to look like a real line)
- Social or platform handles (for example @username)
- Street or mailing addresses
How detection works
When you send a message, we run an automated review that looks for those kinds of patterns and context before the prompt is persisted or passed to the AI.
A secondary pass applies rule-based checks (for example obvious email shapes, @handles, and phone-like digit runs) so obvious leaks are still caught if the primary step misses something.
Placeholders you might see
Sanitized text may contain tokens such as [Person], [Email], [Phone], [Handle], or [Address]. The assistant is instructed to answer helpfully without repeating those tokens or inventing the underlying private facts.
For broader data practices, see our Privacy Policy.
Your privacy beyond this notice
This guide explains prompt redaction in chat. For how we handle cookies, accounts, and overall data use, read our Privacy Policy.