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

OpenAI has introduced a new model called GPT-5.6-Cyber, which is designed for cybersecurity purposes, including vulnerability research, penetration testing, and incident response. This model is based on GPT-5.6 Sol and is trained to enhance capabilities in specialized cybersecurity tasks. It is intended to improve performance in areas such as identifying zero-day vulnerabilities and developing exploit chains. The model has reduced safeguards in place, allowing it to handle higher-risk tasks with fewer refusals. The goal of GPT-5.6-Cyber is to support advanced cybersecurity activities.