OpenAI Launches GPT-5.6-Cyber with Reduced Safeguards for Exploit Development

OpenAI launched GPT-5.6-Cyber, a cybersecurity-focused model for vulnerability research, penetration testing, and incident response. Built on GPT-5.6 Sol, it improves specialized tasks like finding zero-day vulnerabilities and developing exploit chains, with reduced refusals for higher-risk requests. The model aims to enhance security capabilities but raises concerns about potential misuse.