Help·Security 2 min

Safeguard: protect sensitive data

IBANs, names, addresses and passwords are masked before sending — the AI model only sees placeholders.

Safeguard checks every message on our server before it goes to an AI provider. Found data is replaced with placeholders and automatically swapped back in the response.

  1. 1

    Switch on

    The ’Safeguard’ chip in the control bar has three levels: off, automatic (masks without asking) and ask (shows findings first). You set the default in settings.

    Safeguard shows found sensitive data before sending
  2. 2

    What’s detected

    IBAN, card numbers, tax ID, email addresses, phone numbers, addresses, names and passwords. Names are recognised by a local model on our server — no external service.

  3. 3

    For teams

    Main accounts and admins can fix the Safeguard per person. Then it can’t be switched off.

Tips

  • For false positives, such as company names recognised as people, simply deselect the finding in the ask level.
  • PDFs and images aren’t masked, there’s only a warning — redact contracts beforehand.

Security settings

Related guides

Set up team

Sub-users with their own login, limits and models per person, shared projects, one invoice.

EU models and data protection

Which models run in Europe, what EU-only mode does and where the data processing agreement is.