OpenAI Codex Breaks USB Encryption in Minutes, Exposing Hardcoded Keys and Security Flaws

OpenAI Codex helped researcher Xusheng Li uncover flaws in four USB protection products, including hardcoded encryption keys and hidden accounts that allowed files to be decrypted without the owner’s password. In one test, Codex identified the main weakness in seven minutes. The findings expose failures in how these products handle secrets, rather than a breakthrough against encryption algorithms. Xusheng Li published the research on Binary Ninja’s blog on October 22, 2026. Using AI with Binary Ninja’s MCP server, he examined SecZure’s YutianDun, NetacFileLock, KakaSoft USB Copy Protection, and Nexcopy Copy Secure. The work combined software analysis with tests on protected files, showing how quickly AI can help researchers find code that controls encryption. SecZure’s YutianDun hides encrypted files and opens them through readers designed to block copying, screenshots, and printing.