AI safety scanning: what to tell your team
If you turn on AI safety scanning, inform your staff - here is a ready-to-use notice and handbook clause.
For: Owners & admins · Category: Safety & whistleblowing
AI safety scanning checks workplace messages for harassment, threats and data leaks. It is off by default and you enable it per company on the Safety page. Because it analyses staff communications, data-protection law (Nigeria's NDPA and the GDPR) expects you to tell your team before or as it starts. Veridox discloses the feature in its Privacy Policy, but informing your own workforce is your responsibility - use the wording below.
Steps
- What is analysed - In-app chat messages (one-to-one and group), the subject and body of emails sent through a connected mailbox, and certain attendance text. Most messages are classified safe and are not stored.
- What is kept - Only messages the system flags as a possible concern are stored - the flagged text plus the reason, severity and confidence. Routine messages are analysed and immediately discarded.
- Who can see flags - Only HR, admins and the owner can view flagged items, on the Safety page. Treat them as confidential.
- Why you are doing it - To protect staff from harassment and abuse and to protect company and customer data. Your lawful basis is your legitimate interest in a safe workplace.
- Tell your team - Send the notice below (or add the clause to your staff handbook) on or before the day you enable scanning. Do not back-date it.
Tips
- Ready-to-send staff notice: "We have turned on an automated safety check in Veridox. Messages sent through Veridox (chat, and email sent through the platform) are analysed automatically to flag harassment, threats or leaks of confidential information. The check is automated; the vast majority of messages are not stored or read by anyone. Only a message flagged as a possible concern is kept and shown to [owner/HR] for review. This is about catching harm, not monitoring your day-to-day conversations. You can ask what has been flagged about you, ask us to correct it, or complain to the data-protection authority. Questions: [contact]. Effective [date]."
- Handbook clause: "Communications sent through the Company's workplace platform (Veridox), including in-app chat and connected work email, are subject to automated analysis to detect harassment, threats, abuse and unauthorised disclosure of confidential information. Routine communications are not retained; only content flagged as a potential concern is stored and made available to authorised personnel (owner/HR) under access controls. Processing is based on the Company's legitimate interest in a safe workplace and protecting data. Flags are retained for [period] then deleted. Employees may request access to information held about them, seek correction, or complain to the NDPC or relevant authority; contact [contact]."
- This is a transparency notice, not legal advice - have your legal contact confirm the lawful-basis wording and retention period for your jurisdiction before you rely on it.
- On the free plan only about two messages a week are scanned, so coverage is light until you upgrade; the notice is still correct and worth sending.