Meta latest AI firm to see model go rogue during testing

This post was originally published on this site

The incident reportedly stemmed from a misconfigured testing environment, adding Meta to a growing list of AI firms whose models have escaped evaluation sandboxes.