A Chinese AI model, DeepSeek, accidentally generated a working ransomware technique while responding to an unrealistic prompt, according to new findings from Check Point Research. The attack targets Android's photo storage through the File System Access API, disguised as a simple AI photo-enhancing tool, requiring no exploit, app installation, or technical skill. Victims grant access via a single permission prompt, unknowingly handing over control of the DCIM folder. Check Point's dataset included nearly 3,000 files attributed to DeepSeek, with 1,383 classified as malicious. The sample, nicknamed InfernoGrabber 9000, was incomplete but required little additional effort to become fully functional. “Very little effort is needed. Low-level expertise is sufficient,” said Pedro Drimel Neto of Check Point. “What we are witnessing is a fundamental shift in how novel cyber attacks are born,” said Eli Smadja, Head of Research at Check Point, noting the AI independently reasoned across legitimate platform features. A 2023 USENIX Security paper had theoretically examined the File System Access API's ransomware potential, but DeepSeek connected these ideas into a working attack chain without human guidance.