OpenAI has announced a new artificial intelligence model named GPT-5.4-Cyber. Similar to Anthropic's Claude Mythos, this new "cyber-permitted" version is designed for defense cybersecurity and is not for public use.

OpenAI's Latest Version Aims to Pave the Way for More Capable Models

OpenAI states that the new GPT-5.4-Cyber version is specifically designed to pave the way for more capable models coming this year.

In preparation for more capable models coming from OpenAI in the next few months, we are fine-tuning our models specifically to enable defense cybersecurity use cases; as of today, we are starting with a GPT-5.4 version trained to be cyber-permitted: GPT-5.4-Cyber.

Access to GPT-5.4-Cyber is limited to "the highest level users who want to validate themselves as cybersecurity defenders."

Trusted Access Required for Using GPT-5.4-Cyber

OpenAI indicates that the reason for this is that GPT-5.4-Cyber has been "deliberately fine-tuned for extra cyber capabilities and has fewer capability restrictions."

This is a version of GPT-5.4 that provides new capabilities for advanced defense workflows, including binary reverse engineering capabilities that lower the denial threshold for legitimate cybersecurity efforts and allow security professionals to analyze malware potential, vulnerabilities, and security robustness without requiring access to source code.

Since this model is more permitted, we are starting with a limited, iterative deployment for approved security providers, organizations, and researchers.

This deployment is part of an expanded version of the Trusted Access initiative launched by OpenAI earlier this year. The company emphasizes that access to Trusted Access can be obtained through two methods:

  • Individual users can verify their identity at chatgpt.com/cyber.
  • Companies can request trusted access for their teams through OpenAI representatives.

More information about TAC can be found here, and you can learn about GPT-5.4-Cyber.

OpenAI has also introduced a new Pro plan version for Codex users.

  • OpenAI is updating ChatGPT with GPT-5.4 Thinking and offering six significant improvements
  • OpenAI has launched the GPT-5.4 mini and nano, which are its 'most capable small models'
  • OpenAI stated that it is necessary to update Mac applications, including ChatGPT and Codex, as a security measure

Do More with Your Apple Products

Apple AirTag 2 | Add Find My tracking for keys, bags, bikes, and more

Beats USB-A to USB-C Cable | Official CarPlay cable

Wireless CarPlay Adapter | No more need to plug in

AirPods 4 | Apple’s most affordable wireless earbuds

Astropad Rock Paper Pencil | iPad screen protector + Apple Pencil tips

Withings Body Comp | The best smart scale for Apple Health