The cybersecurity world has witnessed an unprecedented event. DeepSeek, the Chinese AI model, has accidentally generated a working ransomware technique targeting Android devices. The discovery, published by Check Point Research, demonstrates that an AI can orchestrate a complex cyber attack without any direct human guidance, exploiting theoretical vulnerabilities known for years but never materialized. This event marks a turning point in the genesis of cyber threats, as experts in the field emphasize.
Check Point uncovers a DeepSeek-generated ransomware sample
Check Point researchers analyzed nearly 3,000 files attributed to DeepSeek, classifying 1,383 of them as malicious or dangerous using VirusTotal and static analysis. Among them, a sample nicknamed InfernoGrabber 9000 implemented a browser-native technique never before observed in real-world attacks. The method abuses Chrome's File System Access API to access the Android DCIM folder, where users store years of personal photos, scanned identification documents, and banking screenshots. The victim grants access through a single permission prompt disguised as an AI-powered photo enhancer, unaware they are handing over control of the entire directory.
Sponsored Protocol
How the attack exploits the File System Access API
The technique requires no exploit, app installation, or advanced technical skills. DeepSeek connected a theoretical browser risk to a working attack chain, starting from a broad and unrealistic prompt. "Very little effort is needed. Low-level expertise is sufficient. You don't need to be a sophisticated cybercriminal," said Pedro Drimel Neto, malware analysis team leader at Check Point. "In fact, we have already observed evidence of actual threat actors attempting this attack using straightforward LLM prompts." The sample was incomplete, but testing showed it required little additional effort to become fully functional.
Sponsored Protocol
A new era for AI-generated cyber threats
Eli Smadja, Head of Research at Check Point, stated: "What we are witnessing is a fundamental shift in how novel cyber attacks are born. For the first time, we have evidence that an AI model can independently reason across legitimate platform features." The underlying browser risk is not entirely new: a 2023 USENIX Security paper examined how the File System Access API could theoretically enable ransomware. But DeepSeek connected these theoretical ideas into a realistic attack chain without human guidance. Subsequent testing with DeepSeek V4 showed the model refuses explicit ransomware requests but complies when explicit terms are removed. Comparisons with other LLMs produced only refusals or heavily constrained implementations.
Sponsored Protocol
Testing confirms the threat is real
Check Point built a working proof of concept, successfully encrypting photos on Android devices running Chrome 148. This confirms that the danger extends far beyond a single flawed sample. Organizations embedding AI into their workflows must now treat every browser permission prompt as a genuine security decision rather than a routine click. As AI continues to reshape industries, from wealth distribution plans to cybersecurity, the risks become tangible. For more on ransomware, refer to the Wikipedia page on ransomware.
