OpenAI expands access to security-focused GPT-5.5 to counter the 'Mythos shock'

\r\n
[Financial News] OpenAI is significantly strengthening its AI support system for cybersecurity.
On the 8th, OpenAI said it will expand Trusted Access for Cyber (TAC) so verified experts can make effective use of the security capabilities of its general-purpose model GPT-5.5, and will begin a preview of GPT-5.5-Cyber for selected partners.
GPT-5.5 is a model with strong capabilities across the entire cybersecurity workflow, including vulnerability discovery and analysis, detection, validation, and patching. The expanded TAC is designed to let verified defenders use AI more flexibly in practical defense work while maintaining the model’s general safety.
With GPT-5.5 under TAC, security teams and developers will be able to efficiently carry out defense-focused workflows within systems they manage, including vulnerability identification and classification, malware analysis, binary reverse engineering, detection engineering, and patch validation. However, requests that could cause real harm, such as attacks on external systems or unauthorized intrusion, will remain strictly restricted as before.
The newly announced GPT-5.5-Cyber is not a model for general release, but a limited preview service for a group of defenders responsible for securing critical infrastructure. Rather than aiming to improve overall cyber performance, it is part of a phased rollout to test user verification, account-level controls, and misuse monitoring systems needed for high-risk defensive work such as red teaming, penetration testing, and vulnerability validation in controlled environments.
OpenAI will also deepen cooperation with the security ecosystem. It plans to work with partners in vulnerability research, software supply chain security, and network security to support the full process from vulnerability discovery to response and network-level mitigation. In particular, to help protect the open-source ecosystem, it will run the Codex for Open Source program and provide major project maintainers with access to Codex Security to support vulnerability identification and remediation.
Through Trusted Access for Cyber (TAC) and the GPT-5.5-Cyber preview, OpenAI plans to gradually expand the use of defender-focused AI in cybersecurity.
Gi-seok Ko, Policy Coordination at OpenAI Korea, said, "OpenAI will continue building a responsible framework for use through trusted access and phased verification, while helping verified defenders detect, analyze, and respond to threats more effectively."
wongood@fnnews.com Reporter Ju Won-gyu Reporter