To AI interception
To AIWhat an employee sees the moment sensitive data is about to leave for an AI tool. To-AI is the only direction Zeflin can block — and it ships warn-first.
Preview the four interception states the extension can render.
Sensitive data detected
To-AI check · not blocked
This appears to contain financial information — are you sure you want to send it to ChatGPT?
We don't show the matched value here. Zeflin only names the category it recognised (financial); the text never leaves your device unredacted.
Blocking is outbound-only, and warn-first.
Only data heading into an AI tool can be blocked, and it ships as a non-blocking warn — hard block is a deliberate admin opt-in. The reverse direction (AI content landing in real work) is monitor, warn, and attest only; it is never hard-blocked.