Critical Zero-Day Found in Widely Used AI Inference Engine

Independent researchers discovered a critical zero-day in a widely used AI inference engine that lets unauthenticated attackers achieve remote code execution on enterprise servers. The vendor released a patch within 48 hours of the October 12 disclosure, but the gap raised concerns about exploitation.

• Flaw in the engine's memory management subsystem allows arbitrary memory writes
• Unauthenticated attackers can gain full control of the host system
• Patch shipped within 48 hours of responsible disclosure on October 12
• AI framework vulnerabilities rose 40% over the past year