Vulnerability Research

An in-depth look at how the safety filters of major LLMs like GPT-4 and Claude 3.5 are creating friction for legitimate security researchers and vulnerability discovery.

OpenAI announces the expansion of its Trusted Access for Cyber initiative, introducing GPT-5.5 and the specialized GPT-5.5-Cyber model to empower verified defenders and protect critical infrastructure.

A groundbreaking partnership between Mozilla and Anthropic reveals how Claude 3.5 Sonnet detected 22 vulnerabilities in the Firefox codebase, including 14 high-severity bugs, marking a new era for AI-driven cybersecurity.