What happened
OpenAI has introduced GPT-6 Cyber, a specialized AI model designed to identify, validate, and help remediate software vulnerabilities. The model delivers substantial improvements over its predecessor, GPT-5.6 Cyber, including a 51% relative performance increase on the ExploitGym cybersecurity . GPT-6 Cyber achieved a 48.5% success rate on ExploitGym, compared with 32.1% for GPT-5.6 Cyber. The model also demonstrated improved performance in vulnerability discovery, identifying 40.3% of vulnerabilities in web penetration testing without source code access, compared with 32% for GPT-6 Astra.
GPT-6 Cyber achieved a 48.5% success rate on ExploitGym, compared with 32.1% for GPT-5.6 Cyber.
The model also demonstrated improved performance in vulnerability discovery, identifying 40.3% of vulnerabilities in web penetration testing without source code access, compared with 32% for GPT-6 Astra.
GPT-6 Cyber successfully completed 95% of tasks on RemoteExploitBench, a involving known vulnerabilities in applications including GitLab, PostgreSQL, and Jenkins.
The model avoided exploiting out-of-scope services in 99.93% of tests without deployment-specific safety instructions, compared with 94.49% for GPT-5.6 Cyber.
With deployment safety instructions enabled, GPT-6 Cyber remained within authorized boundaries in 100% of the evaluated cases.
Why it matters
The launch of GPT-6 Cyber represents a significant expansion of OpenAI’s efforts to develop advanced AI systems for cybersecurity research, vulnerability discovery, and defensive security operations. The model’s improved performance and capabilities are intended to help security professionals use advanced offensive research capabilities without allowing AI agents to expand their activities beyond permitted systems. OpenAI has also demonstrated the model’s effectiveness in real-world software security research, identifying previously unknown vulnerabilities in the Firefox browser and Linux kernel.
The launch of GPT-6 Cyber represents a significant expansion of OpenAI’s efforts to develop advanced AI systems for cybersecurity research, vulnerability discovery, and defensive security operations.
The model’s improved performance and capabilities are intended to help security professionals use advanced offensive research capabilities without allowing AI agents to expand their activities beyond permitted systems.
OpenAI has also demonstrated the model’s effectiveness in real-world software security research, identifying previously unknown vulnerabilities in the Firefox browser and Linux kernel.
The model’s capabilities are intended to help security professionals strengthen critical software infrastructure and accelerate vulnerability discovery, validation, and remediation.
OpenAI is positioning GPT-6 Cyber as a tool for accelerating vulnerability discovery, validating security weaknesses, and helping defenders strengthen critical software infrastructure.
Interactive Mechanism: How It Actually Works
Explore the underlying technology behind this development interactively.
crm_get_transaction(id='4092').A route planner searches possible journeys using explicit rules. What does this illustrate about AI?
What to watch next
The launch of GPT-6 Cyber and its potential impact on the cybersecurity industry. The model’s improved performance and capabilities are intended to help security professionals use advanced offensive research capabilities without allowing AI agents to expand their activities beyond permitted systems.
The potential impact of GPT-6 Cyber on the cybersecurity industry and its potential applications in real-world software security research.
The model’s improved performance and capabilities, including its ability to identify and validate software vulnerabilities.
The potential risks associated with increasingly capable autonomous cybersecurity systems and the need for safeguards to prevent unauthorized activities.
The expansion of OpenAI’s Daybreak cybersecurity program, which provides different levels of model access based on security professionals’ needs and authorization.
The company’s $1 billion Daybreak for Frontline Defenders grant program, which will subsidize access to advanced security technology for resource-constrained organizations.