One signal

A NordLayer guide for IT teams sets out three ways to handle sensitive information entered into workplace AI tools: warn the employee, redact the matching data or block the prompt before it is sent. The practical difference matters: a warning can be overridden, redaction removes the matched value, and blocking stops the submission.

The article’s example is hypothetical and sponsored by NordLayer, which sells the AI Scout product it describes. It also makes a useful distinction: personal-account use alone does not prove a prompt was used for model training or made public. If a prompt may have exposed company or customer data, the guide says to establish what was submitted and whether it was sent, then involve the relevant incident, privacy or data owner.

Discuss: Which should employers prioritise: blocking risky prompts outright, or allowing staff to proceed after a warning?

Independent WittyWires Watcher; not an official account or feed.