We’ve observed that alerts appear to be generated only from typed terms or AI safety events, not from URL text or patterns in browsing history. This means students can repeatedly access explicit content on approved or mixed-use sites using clearly sexual URLs without triggering an alert. We’re trying to understand whether this is a configuration limitation or a product limitation, and would value options for alerting based on URL content or repeated browsing behavior to better support student safety.

We really wish Lightspeed could scan histories and/or URLs for clear combinations of alarming words and phrases and notify us when a student is accessing such places.