OpenAI has officially introduced GPT-5.6-Cyber, a new Artificial Intelligence (AI) model tailored specifically for the cybersecurity industry. This model focuses on assisting security professionals with vulnerability research, penetration testing, and incident response.

Built on the GPT-5.6 Sol foundation, this Cyber version features reduced safeguards. This means the AI is permitted to process high-risk commands and dual-use cyber tasks that standard public AI models typically refuse.

Capabilities and Critical Bug Discoveries

To measure its capabilities, OpenAI conducted internal testing. The results were highly impressive:

• GPT-5.6-Cyber successfully completed 95% of advanced cybersecurity requests. This significantly outperforms the standard GPT-5.6 Sol, which only completed 1.5% of similar tasks.
• The AI has proven effective in finding dangerous bugs. One of its achievements includes detecting CVE-2026-15903, a critical flaw in Google’s V8 JavaScript engine, alongside hundreds of vulnerabilities in mobile operating systems and popular OS kernels.

Restricted Access and AI Challenges

Given the potential danger if the technology falls into the hands of hackers, OpenAI restricts access to GPT-5.6-Cyber. It is only available to trusted corporate partners (such as Cisco, CrowdStrike, and Cloudflare) through a specialized access tier called Daybreak Red.

While AI has become incredibly adept at finding security flaws, its ability to fix those bugs is still lacking. According to research, the success rate of AI generating a perfect patch without altering other system behaviors is only around 26%. Therefore, strict oversight from human security experts remains an absolute necessity.

Source: https://thehackernews.com/2026/08/openai-launches-gpt-56-cyber-with.html