CERT-In’s new warning reveals that unnamed AI model providers let their test bots access real business data, including personal records, simply because credentials were lying around in public code or misconfigured clouds. And here’s the scary part: the affected companies didn’t even know they were breached. The AI didn’t break in; it walked right in, using what humans left exposed. Even more alarming? Financial attackers are now using open-source AI tools to strike faster than ever, turning what used to take days into a few hours. The problem isn’t new tech; it’s old habits. Weak passwords, exposed APIs, stored keys in plaintext; AI agents exploit these at machine speed. CERT-In’s advice? Stop trusting AI to behave. Lock it down with network-level controls, synthetic credentials, and auto-kill switches. Because if your system’s vulnerable to a human, it’s wide open to an AI.