OpenAI Challenges Anthropic with GPT-5.4-Cyber for Security

OpenAI Introduces GPT-5.4-Cyber to Enhance Cybersecurity

OpenAI has recently unveiled GPT-5.4-Cyber, a specialized version of its artificial intelligence model designed to meet the needs of defensive cybersecurity. This model is capable of performing complex tasks such as binary reverse engineering, which involves analyzing compiled software without access to the source code. For now, access to this model is restricted to a few hundred carefully vetted security professionals.

This strategic launch places OpenAI in direct competition with Claude Mythos from Anthropic, an AI model launched a week earlier that focuses on detecting vulnerabilities in operating systems and browsers. Like GPT-5.4-Cyber, Claude Mythos is also limited to a select group of users.

A More Permissive Model but Under Control

OpenAI's "Trusted Access for Cyber" (TAC) program is expanding with the introduction of GPT-5.4-Cyber. This model is designed to be less restrictive regarding defensive security, but this permissiveness comes with rigorous controls. Access via third-party platforms or under data retention agreements may be limited to ensure security.

OpenAI plans to extend access to thousands of verified individuals and hundreds of teams in the coming weeks, further reinforcing its commitment to digital security.

Security Initiatives and Investments

Alongside the launch of GPT-5.4-Cyber, OpenAI is highlighting its Codex Security tool, which has already helped fix over 3,000 critical vulnerabilities. This tool automatically identifies flaws in codebases and suggests corrective solutions.

OpenAI is also investing $10 million in a grant program for cybersecurity projects, having already reached over 1,000 open-source projects with a free security scan.

Reactions and Concerns

Anthropic's Claude Mythos model has raised concerns among financial firms and U.S. government agencies. During a meeting with Wall Street executives, Treasury Secretary Scott Bessent and Fed Chair Jerome Powell emphasized the need to take this model seriously. The Treasury Department is seeking access to Mythos to assess its own security systems.