Unit 42 Continuous Frontier AI Defense uses always-on testing to detect network vulnerabilities before hackers exploit them, as cybersecurity firms race to counter increasingly automated cyberattack methods.
Palo Alto Networks has unveiled a new enterprise cybersecurity offering built on artificial intelligence models developed by Anthropic and OpenAI, designed to detect and close network security gaps before attackers can take advantage of them.
The move reflects a broader shift in the cybersecurity industry, as hackers increasingly turn to automated tools to scan for and exploit weaknesses in corporate systems. Security firms are now under pressure to build defenses that can operate at a similar speed and scale.
Continuous testing replaces periodic scans
The new platform, called Unit 42 Continuous Frontier AI Defense, combines open-weight AI models with cybersecurity-focused systems, including offerings from both Anthropic and OpenAI. Rather than relying on scheduled vulnerability scans, the service runs continuously, adapting to changes within an organization's digital environment in real time.
The system monitors web applications, APIs, and cloud infrastructure on an ongoing basis, tracing full attack paths so security teams can identify not just isolated flaws but the complete routes an intruder might use to breach a network.
Palo Alto Networks said the shift toward automation on the offensive side of cybersecurity has dramatically shortened the time attackers need to identify and exploit weaknesses, in some cases cutting the process from weeks to just hours. The company positioned its new platform as a way for businesses to test whether their defenses can keep pace with that accelerated threat environment.
Guided fixes and early results from testing
Beyond flagging vulnerabilities, the platform is designed to provide practical remediation support, including specific code-level fixes and temporary protective measures that can shield systems while permanent repairs are put in place.
The company said the platform was developed over six months of internal testing and validated across more than 100 customer engagements, supported by an investment of $17 million in research and development. During its own internal rollout, Palo Alto Networks reported uncovering a year's worth of security exposures in just three weeks.
Across customer evaluations, the company said vulnerabilities were found in every organization tested, with more than a third of those classified as high or critical severity. A significant share of the exposures were traced to internally developed applications, while the majority of issues found in third-party software had no previously documented vulnerability record.
Palo Alto Networks confirmed the service will roll out globally through an annual subscription model, with pricing varying based on which combination of AI models — from OpenAI, Anthropic, or open-source providers — a customer selects for deployment.
