OpenAI Locks GPT-5.4-Cyber for Verified Experts

GPT-5.4-Cyber: An AI Tailored for Cybersecurity

OpenAI's GPT-5.4-Cyber model is not an entirely new creation but rather an adaptation of the GPT-5.4 model, specifically oriented towards cybersecurity. OpenAI has announced that this version has been optimized to meet the needs of cybersecurity professionals. Two major improvements have been made:

• Advanced Cyber Capabilities: GPT-5.4-Cyber is equipped to perform complex cybersecurity tasks, such as binary reverse engineering. This allows experts to analyze compiled software to detect potential malware, security vulnerabilities, and assess the robustness of a system without requiring access to the source code.

• Fewer Restrictions: This model reduces the limitations that often prevent AIs from responding to certain requests due to fears of misuse. GPT-5.4-Cyber is designed to continue functioning in scenarios where other models might refuse to execute legitimate cybersecurity tasks.

These features explain why OpenAI has not made GPT-5.4-Cyber available to the general public.

Why is GPT-5.4-Cyber Restricted?

OpenAI has chosen not to immediately open access to GPT-5.4-Cyber to prevent its capabilities from being misused. The model is integrated into the Trusted Access for Cyber (TAC) framework, a system based on trust and user identification. This framework aims to ensure that only verified individuals, who have proven their legitimacy as cybersecurity defenders, can use these powerful tools.

OpenAI has begun to expand TAC to a larger number of professionals, including thousands of individual defenders and hundreds of teams dedicated to protecting critical software. The goal is to allow these verified users to access the advanced features of models GPT-5.2 to GPT-5.4 while maintaining strict control over their usage.

OpenAI has clearly expressed its concern: AI is evolving rapidly, and if misused, it can pose a danger to cybersecurity. Therefore, the maximum capabilities of AI in the field of cybersecurity must be carefully controlled and kept in good hands.

Who Can Access GPT-5.4-Cyber?

Access to GPT-5.4-Cyber is strictly reserved for existing TAC clients. These users must be prepared to prove their status as legitimate defenders in the field of cybersecurity to be eligible. OpenAI has established several levels of access for its enhanced models, each designed to meet specific security needs.

These models are capable of handling critical security tasks that general AIs might refuse, such as security education, defensive programming, and vulnerability research. For those at the top of the TAC pyramid, OpenAI provides precise instructions for obtaining access to GPT-5.4-Cyber.

How to Obtain Access to GPT-5.4-Cyber?

To access GPT-5.4-Cyber, users must follow a specific process:

• Registration for TAC: Individuals must verify their identity via the website chatgpt.com/cyber, while companies must request trusted access through their OpenAI representative.

• Access Request: Once registered, users can request access to GPT-5.4-Cyber. However, approval is not guaranteed and depends on validation by OpenAI.

Anyone wishing to gain access will have to undergo a rigorous identification process. Ultimately, this strict control could transform the way we secure cyberspace, providing defenders with the most advanced tools available while staying one step ahead of potential cybercriminals.