OpenAI releases GPT-5.4-Cyber, a cybersecurity-focused model limited to verified security professionals
OpenAI has released GPT-5.4-Cyber, a fine-tuned variant of GPT-5.4 built for defensive cybersecurity work including binary reverse engineering. Access is initially restricted to a few hundred verified security professionals, with expansion planned to thousands of individuals and hundreds of teams in coming weeks.
OpenAI releases GPT-5.4-Cyber, a cybersecurity-focused model limited to verified security professionals
OpenAI has released GPT-5.4-Cyber, a fine-tuned variant of GPT-5.4 built for defensive cybersecurity work including binary reverse engineering. Access is initially restricted to a few hundred verified security professionals.
Access restrictions and rollout
The model is being distributed through OpenAI's "Trusted Access for Cyber" (TAC) program. A few hundred users have initial access, with OpenAI planning to expand to thousands of verified individuals and hundreds of teams over the coming weeks.
According to OpenAI, GPT-5.4-Cyber is less restrictive than standard models when performing defensive security tasks. The company specifically highlights binary reverse engineering — analyzing compiled software without source code access — as a key capability.
Because of the model's more permissive nature, OpenAI has implemented tighter access controls. Access through third-party platforms or under zero-data-retention agreements may be restricted.
Direct competition with Anthropic
The release comes one week after Anthropic unveiled Claude Mythos, an AI model specialized in finding and exploiting vulnerabilities in operating systems and browsers. Like GPT-5.4-Cyber, Mythos is only available to a restricted user group.
According to Bloomberg, Mythos has already raised concerns among financial firms and U.S. government agencies. Treasury Secretary Scott Bessent and Fed Chair Jerome Powell reportedly warned Wall Street executives to take the model seriously during a meeting. The Treasury Department's technology team is attempting to obtain access to Mythos to check its own systems for vulnerabilities.
Broader cybersecurity push
Beyond the new model, OpenAI points to its existing Codex Security tool, which automatically identifies vulnerabilities in codebases and suggests fixes. The company claims the system has helped patch more than 3,000 critical vulnerabilities since launch.
OpenAI is allocating $10 million to its Cybersecurity Grant Program and reports reaching over 1,000 open-source projects with free security scanning.
What this means
The near-simultaneous releases of GPT-5.4-Cyber and Claude Mythos signal a new competitive front in AI development: specialized models for security work. The restricted access approach reflects industry awareness that these capabilities could be weaponized if widely available. The Treasury Department's scramble to access Mythos for defensive purposes shows how quickly government agencies are moving to assess AI-powered security tools — both as threats and potential defenses.
Related Articles
Moonshot AI releases Kimi K3, China's largest model at 2.8 trillion parameters
Beijing-based Moonshot AI released Kimi K3, China's largest AI model at 2.8 trillion parameters. The company claims the model consistently outperforms OpenAI's GPT 5.5 and Anthropic's Claude Opus 4.8 on benchmarks including coding and general agents, though it still trails the leading-edge GPT 5.6 Sol and Claude Fable 5 in overall performance.
OpenAI's GPT-5.6 Codex Bug Deletes User Files When Attempting to Override $HOME Environment Variable
OpenAI has identified a critical bug in GPT-5.6's Codex implementation that causes unexpected file deletions. According to Thibault Sottiaux, the issue occurs when the model attempts to override the $HOME environment variable to define a temporary directory but mistakenly deletes $HOME instead, particularly when full access mode is enabled without sandboxing protections.
OpenAI GPT-5.6 Sol, Terra, and Luna launch on Amazon Bedrock with 80-point Coding Agent Index score
OpenAI's GPT-5.6 model family is now generally available on Amazon Bedrock, introducing a three-tier system: Sol (flagship reasoning), Terra (balanced production), and Luna (fast inference). According to OpenAI, Sol scores 80 points on the Artificial Analysis Coding Agent Index and 73.5% on ExploitBench, establishing new benchmarks while using less than half the output tokens of competing models.
OpenAI's GPT-5.6 Sol Adds Five Reasoning Effort Settings, Follows DeepSeep-R1 RLVR Training Method
OpenAI released GPT-5.6 Sol, a new reasoning model family that comes in three sizes with roughly five to six reasoning-effort settings each. The release follows the DeepSeek-R1 methodology of using reinforcement learning with verifiable rewards (RLVR), nearly two years after OpenAI's original o1 model popularized LLM-based reasoning.
Comments
Loading...