OpenAI launches GPT-6 Cyber model for cybersecurity research

OpenAI has introduced a new AI model named GPT-6 Cyber to assist professionals in the field of cybersecurity. This model is specifically designed to identify software vulnerabilities and develop exploits. The company stated that to ensure its capabilities are not misused for cyberattacks, the model will initially be available only to a limited number of approved security customers.

According to a report by News18, GPT-6 Cyber outperforms the company's general-purpose GPT-6 Astra model in most cybersecurity tasks. As per reports from Fortune and TechCrunch, this model focuses on advanced vulnerability research and exploit development, which will help security researchers find software flaws and investigate how attackers might exploit them.

OpenAI researchers utilised GPT-6 Cyber on Mozilla Firefox and the Linux kernel to test its capabilities. The company stated that the model helped discover 81 previously unknown high-severity vulnerabilities in Firefox. Researchers also developed working exploits for 948 Linux kernel vulnerabilities, including 588 previously unknown flaws. According to OpenAI, some of these Linux exploits could allow an attacker with common local access to gain the highest level of control over the system.

Compared to a study conducted in July using GPT-5.6 Sol, researchers developed more than five times as many working exploits for Linux vulnerabilities. This launch comes at a time when concerns regarding the use of AI in cyberattacks are increasing. Gartner estimated in September that global spending on AI cybersecurity would reach $51.3 billion this year, up from $25.9 billion last year.

OpenAI is introducing its fourth model focusing on cybersecurity this year, following similar releases in April, June, and August. The company is expanding its enterprise business in the cybersecurity sector and recently appointed Dali Rajic, who has experience in cybersecurity software, as Chief Revenue Officer. Through its Daybreak programme, the company currently offers two tiers of access to trusted professionals: Daybreak Red for advanced testing such as attack simulation, and Daybreak Blue for defensive tasks such as code review and malware analysis.

OpenAI has warned that cybersecurity professionals have limited time to strengthen digital infrastructure before AI models become even more capable of finding software vulnerabilities. OpenAI faces competition in this market from Google and Anthropic. Anthropic has introduced its Glasswing programme and Google its Fairwind programme, allowing selected security professionals to use their models without certain restrictions.

OpenAI is an artificial intelligence research and deployment company known for developing various types of AI models.