[Digital Today reporter Chi-gyu Hwang (황치규)] Zscaler is working with OpenAI to apply the latest AI models across its security platform.

Zscaler has joined OpenAI's Trusted Access for Cyber (TAC) programme, gaining access to security-focused models including GPT-5.4-Cyber, which is tailored for defensive cybersecurity, and related application programming interfaces (APIs).

Zscaler plans to integrate the models into its Zero Trust Exchange platform, AI red team service and security operations service.

TAC is a programme that gradually allows access to high-performance models only to vetted security companies. GPT-5.4-Cyber is optimised for defensive tasks such as vulnerability discovery, binary analysis and attack-chain inference.

Zscaler will integrate GPT-5.4-Cyber and Codex-based security models into its multi-agent security architecture and software development life cycle (SDLC). Developers can receive AI-based security reviews at the code-writing stage. Zscaler aims to find vulnerabilities before deployment and reduce exposure time.

For customers, it is pushing two functions forward. The first is its AI red team service. Using OpenAI text, image and voice models, it automatically generates various attack scenarios such as prompt injection, jailbreaks and model confusion, and also presents immediate response measures.

The second is its managed detection and response (MDR) service. Zscaler operates the MDR service by combining AI agents and human analysts through the Red Canary MDR model.

AI handles alert classification, signal correlation analysis and timeline construction, while experts make the final judgment. Zscaler emphasised that it is maintaining detection accuracy of 99.6 percent through this.

Earlier, TrendAI, the enterprise cybersecurity business unit of Trend Micro, also formed a strategic partnership with Anthropic. Through the collaboration, TrendAI will embed Claude models across its platform to strengthen agentic workflows, automation and AI-native security operations, and push threat research to identify vulnerabilities in AI systems and infrastructure.