Skip to main content
We can mask personal data out of your prompt before the provider sees it. Each match becomes a stable placeholder such as [EMAIL_1], so the model can still tell people apart. It covers emails, phone numbers, and patterns such as card numbers.
Masking runs on Managed Keys. On your own keys we forward your request untouched. If your request sets pii_mask we follow it. If it doesn’t, we follow your organization’s default. If neither is set, masking is off. That means you can send false to turn masking off for one request when your organization has it on. If masking is on and it fails, we stop the request rather than send it unmasked.