Palo Alto Networks and OpenAI reveal breakthrough cyber model
The expanding collaboration between the two giants will for the first time allow commercial access to an advanced cyber model, aimed at simulating attacks and exposing organizational vulnerabilities before hackers exploit them.
The collaboration between cyber giant Palo Alto Networks and OpenAI is expanding, with the goal of introducing advanced artificial intelligence capabilities into the defensive side of the cyber world. As part of this move, security teams in organizations will gain access to the GPT 5.6 Daybreak cyber model from OpenAI, which until now had not been available for commercial use.
The central idea behind this initiative is to allow cyber defense systems to use artificial intelligence proactively, attempting to think and act as an attacker would. The system can simulate attacks within the organizational environment, actively search for penetration methods, and identify attack paths that might remain undiscovered using conventional security tools.
The need for such capabilities is highlighted by data presented by Palo Alto Networks researchers. According to the company, 36% of security exposures detected were not related to known vulnerabilities with an official CVE identifier. This means that relying solely on databases of known vulnerabilities is insufficient to deal with evolving threats, especially in an era when cyber attackers are also using advanced AI tools.
The Unit 42 research unit of Palo Alto Networks is expanding the Frontier AI Defense service as part of this move. The system is capable of detecting vulnerabilities, incorrect security settings, leaked access credentials, and various assets, and then testing how they can be connected into a full attack path.
"In recent months, the discourse around advanced AI models has focused mainly on the threat they pose to cyber defense teams," said Sam Rubin, Senior Vice President of the Unit 42 research unit at Palo Alto Networks.
According to him, these same capabilities can now be directed to the defensive side. "We are transferring these capabilities to the side of the defenders: not just the knowledge we learned from the models, but the models themselves, which operate within the organization's environment to locate and test attack methods before the attackers exploit them," explained Rubin.
This move marks a shift in the perception of cyber defense, from relying on identifying known threats to attempting to anticipate how an attack could develop. Instead of settling for periodic vulnerability scans, the technology allows for performing attack simulations on an ongoing basis and searching for combinations between different vulnerabilities.
