OpenAI built a model it doesn’t want most people to use
OpenAI on Monday released GPT-5.6 Cyber, a model trained for security work that general-purpose models routinely block. It’s now available The post OpenAI built a model it doesn’t want most people to use appeared first on The New Stack .
OpenAI launched GPT-5.6 Cyber, a specialized security model, alongside GPT-5.6 Sol. While GPT-5.6 Cyber excels at tasks like exploit chain creation, authentication bypass, and privilege escalation, it does not perform as well on overall performance. GPT-5.6 Cyber achieved 95% success in security tests, while GPT-5.6 Sol only achieved 1.5% without standard safeguards and 2% with lower restrictions.
GPT-5.6 Cyber is designed for advanced security tasks like zero-day discovery and exploit chain building, while GPT-5.6 Sol is suited for secure code review, malware analysis, incident response, and vulnerability discovery. Both models have different access tiers and limitations. OpenAI has integrated GPT-5.6 Cyber into products like Palo Alto Networks' Frontier AI Defense and SentinelOne's Wayfinder Frontier AI Services.
GPT-5.6 Cyber demonstrated its effectiveness by uncovering previously unknown vulnerabilities in widely-used software, but it still requires careful handling and monitoring to ensure secure usage.
Written by urgent.news from The New Stack's reporting — not their text. Machine-written — it may contain errors, so check the original before relying on it.
