Meta latest AI firm to see model go rogue during testing

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

Read at source