OpenAI Releases GPT-5.4-Cyber, AI Specially Designed to Find Cybersecurity Vulnerabilities

The company behind the ChatGPT chatbot, OpenAI, introduced its latest AI model named GPT-5.4-Cyber on Tuesday (14/4/2025). This is a flagship AI model specifically focused on cybersecurity needs.

This AI model serves as a direct competitor to the AI from Anthropic, which was launched last week, called Mythos.

In an official blog post, OpenAI explained that GPT-5.4-Cyber is designed specifically to assist professionals in the cybersecurity field in detecting and analysing potential system vulnerabilities.

OpenAI stated that this model is actually a version of GPT-5.4 that has been fine-tuned to be more reliable, agile, and responsive in cybersecurity tasks.

One of the highlighted capabilities of GPT-5.4-Cyber is Binary Reverse Engineering. This is a technique for analysing software in its compiled form.

Although considered quite advanced for being able to detect risks from finished products, OpenAI notes that this AI model is sensitive and has the potential to be misused by irresponsible individuals, for example, to hack systems.

Therefore, OpenAI is currently providing initial access to GPT-5.4-Cyber on a limited basis to verified cybersecurity vendors, organisations, and researchers.

There is no information yet on whether this AI model will be widely available to the public.

Alongside the launch of GPT-5.4-Cyber, OpenAI is also expanding the Trusted Access for Cyber (TAC) programme, which was introduced in February 2026.

Through this programme, OpenAI aims to enable thousands of individuals and hundreds of teams responsible for protecting critical software to access AI technology gradually.

To obtain access, individual users must go through an identity verification process, while companies can apply through OpenAI representatives, as summarised by KompasTekno from OpenAI.