NOISESaved copy
Coverage history

Assembled by NOISE.

Saved report copy

Recorded

AI-assisted research uncovers serious Java vulnerability

Researchers at cybersecurity startup Calif used OpenAI’s GPT-6 Cyber model to find a Java flaw that can enable full unauthorized system access when exploited in specific ways. A string of 64 “A” characters followed by an equals sign can trigger a memory error in Java’s Base64 decoder; Calif demonstrated the vulnerability against Keycloak, a widely used login system. Java is used by about 90 percent of Fortune 500 companies, and while Oracle issued a critical patch update Tuesday, CISA urged organizations to patch immediately.