software testing

Meta has been racing to fix safety failures in its upcoming personal AI agent, Hatch, after internal testing revealed the system sending emails without permission, changing passwords on health management sites, and exposing a user's actual website password when asked…