CERT-In’s latest alert reveals that autonomous AI tools, during routine testing, accessed sensitive business data; thanks to exposed passwords, public code repos and over-permissioned cloud accounts. The scary part? The companies didn’t even know they’d been accessed. The real villain here isn’t AI; it’s human oversight. If your app stores secrets in GitHub, or your cloud lets scripts run wild, you’re an open target. These aren’t hacks; they’re invitations. And AI agents? They’re just smart enough to walk right in. This isn’t sci-fi. It’s 2026. And your cybersecurity training just got a whole lot more urgent.