OpenAI Launches GPT-5.4 Cyber Model for Defensive Cybersecurity To Rival Anthropic's Claude Mythos Preview AI model

Hyderabad: OpenAI has launched a specialised variant of its GPT-5.4 model, GPT-5.4 Cyber, which is designed specifically for defensive cybersecurity applications. Unlike standard consumer-facing models, GPT-5.4 Cyber will not be available on ChatGPT. Instead, it will be rolled out exclusively to vetted security vendors, organisations, and researchers. It is worth noting that the GPT-5.4 Cyber has been launched just a week after rival Anthropic announced its Claude Mythos Preview AI model under Project Glasswing, which is designed for defensive cybersecurity purposes.

What Is GPT-5.4 Cyber?

OpenAI described GPT-5.4 Cyber as an AI model designed to lower the refusal boundary for legitimate security work. Standard models are equipped with strict guardrails that can prevent professional cybersecurity tasks, while GPT-5.4 Cyber is explicitly built to address that limitation.

The key feature of the new AI model is binary reverse engineering. This enables security professionals to analyse compiled software for malware, vulnerabilities, and overall security robustness without requiring access to the original source code. OpenAI stated that its goal is to make advanced defensive capabilities available to those responsible for protecting critical infrastructure, public services, and essential digital systems.