AI applications' default connection protocol, MCP, has four attack surfaces; here's how to secure them.

New malware, JarService/zhima, targets connected vehicles by compromising Android car head units through their update mechanisms.

Western AI labs agree to watermarking, but China's open-source community may resist EU pressure, creating a global AI governance divide.

A college student's sharp observation exposed a sophisticated AI-driven attempt to compromise a major tech platform, highlighting new frontiers in cybersecurity threats.

A deep dive into the critical components of a server security audit, from attack surface mapping to intrusion detection.

Don't trust your LLM endpoint to tell you what model it is. Build a verification harness instead.

A novel attack uses encrypted payloads to bypass LLM defenses, demonstrating a new threat vector for AI agents.

Weak access controls on Windows named pipes expose privileged services. ThreatLocker outlines essential security measures.

US authorities warn of sophisticated attacks on Siemens S7 PLCs using AI to generate exploits, risking disruption to water and other vital services.

Rockstar's parent company is demanding Windows device IDs from Microsoft to unmask GTA 6 leakers across three Discord servers.