Competition in the field of artificial intelligence (AI) continues to heat up when OpenAI announces the new GPT-5.4-Cyber model, just a week after Anthropic launched the Mythos system for cybersecurity.
GPT-5.4-Cyber is a refined variant from OpenAI's advanced model line, focusing on supporting cybersecurity defense.
According to the announcement, this tool is designed to help experts detect, analyze and handle security vulnerabilities in software systems.
This move comes right after Anthropic introduced Mythos within the framework of the "Glasswing Project". This model is currently being tested in a controlled manner, allowing some organizations to use the unreleased version for defensive purposes.
According to Anthropic, Mythos has discovered thousands of serious vulnerabilities in operating systems, browsers and many other software.
Faced with competitive pressure, OpenAI chose a more cautious approach. GPT-5.4-Cyber will not be widely available immediately but will only be deployed limitedly to security providers, organizations and censored researchers.
The company said that access control aims to minimize the risk of technology abuse in cyber attacks.
Along with that, OpenAI expands the Trusted Access for Cyber (TAC) program, which was launched earlier this year.
This program provides access to multiple levels, in which users who are verified at the highest level will be able to use GPT-5.4-Cyber with fewer restrictions for sensitive tasks such as vulnerability research.
Experts believe that the emergence of specialized AI models for network security shows a new trend when AI is not only a tool for creating content but also becomes a "digital shield" against increasingly sophisticated threats.
However, balancing technology power and risk control is still a big problem for AI development companies.
