Study: Potentially dangerous AI-generated proteins are not always recognised

A study shows that AI can create dangerous proteins that bypass common security checks. Software patches should help.