Red-team testing
noun1.It's when a company hires smart people to deliberately try to break or hack their AI system before it goes public - like paying someone to find the holes in your fence before a burglar does. Claude's handlers did exactly this and found real security gaps. It's a safety check, not a failure.
Straight up:“Testing your AI's defences by hiring people to attack it. Sounds like a fight club, works like good sense.”The Straight-Shooter
Explained 2 August 2026 on the 6 O’Clock News.