Calif uses OpenAI's GPT-6 Cyber model to find Java flaw triggered by 64 'A' characters
AI found a hole in Java, and it starts with 'AAAAAAAA '
Security startup Calif found a Java bug with help from OpenAI's GPT-6 Cyber model, which is not yet publicly available: feeding the base64 decoder a string of 64 'A' characters followed by an equals sign triggers a memory error that can crash the system or, when finessed, grant full unauthorized access. Calif demonstrated the flaw against Keycloak, a login system used by organizations including CERN and NASA, and said it is not easy to weaponize. Oracle shipped a critical patch update on Tuesday without specifically mentioning the bug, and CISA urged organizations to patch immediately; Calif estimates the bug could linger for five to 10 years.
A concrete case of an AI model surfacing a Java memory flaw that can grant full system access, showing both the reach of automated vulnerability discovery and the patching gap it exposes.
Source:Hacker News · AI · engadget.com