OpenAI Launches Cybersecurity AI Model for Select Users
OpenAI has introduced a new artificial intelligence model aimed at detecting software security vulnerabilities, granting access to a limited group of users. The move comes just one week after rival Anthropic PBC unveiled its own cybersecurity tool, Mythos, intensifying competition in the AI security space.
GPT-5.4-Cyber Designed for Defensive Security Use
The San Francisco-based company began rolling out its new model, GPT-5.4-Cyber, on Tuesday. The system is specifically developed to identify and address digital weaknesses.
OpenAI stated that it is refining its models to support defensive cybersecurity applications, introducing a version of GPT-5.4 that is optimized for cyber-permissive use cases. This allows security professionals to more effectively test and analyze vulnerabilities.
Trusted Access Program Enables Controlled Testing
Currently, GPT-5.4-Cyber is available to participants in OpenAI’s Trusted Access for Cyber program, launched in February. The initiative gives cybersecurity experts the ability to experiment with advanced AI tools under fewer restrictions, enabling deeper vulnerability testing.
Mythos Sparks Concern Among Financial and Government Leaders
Anthropic’s competing Mythos model has already raised concerns among major financial institutions and government authorities. According to reports, U.S. Treasury Secretary Scott Bessent and Federal Reserve Chair Jerome Powell recently warned Wall Street executives about the potential risks linked to the technology.
Officials emphasized that organizations should take these emerging AI-driven threats seriously, especially as digital infrastructure has long been vulnerable even before the rise of advanced artificial intelligence.
Rising AI Threats and Expanding Access
OpenAI highlighted that threat actors are actively exploring new AI-powered methods to exploit systems. In response, the company plans to expand its program from a few hundred testers to thousands of verified cybersecurity professionals in the coming weeks.
At the same time, OpenAI reaffirmed its commitment to broader access, aiming to democratize the use of advanced tools while maintaining strict identity verification standards for users.
