As attackers leverage artificial intelligence to accelerate their exploitation of security flaws, Palo Alto Networks is fighting back with its own machine-powered defense. The company's threat research division, Unit 42, unveiled Continuous Frontier AI Defense today—a subscription offering that deploys tightly controlled AI models to continuously probe customer infrastructure for weaknesses.
The service runs on two purpose-built cybersecurity models with restricted access: Anthropic's Claude Mythos 5 and OpenAI's GPT-5.6-Cyber, supplemented by open-weight alternatives. Unit 42's proprietary software orchestrates which model tackles each specific task, a design choice the company says optimizes both detection quality and operational costs.
Claude Mythos found flaws that survived decades of human review, and more than ten thousand high-severity vulnerabilities across the software the world runs on.
Michael Moore, Cybersecurity Lead, Anthropic
The offering extends beyond a single-point assessment. Following an initial comprehensive scan of a customer's environment, the system continuously attacks web applications, APIs, cloud infrastructure, code repositories, and network systems as they evolve. For each identified weakness, the service attempts to validate exploitability, traces potential lateral movement paths, and delivers either code-level fixes or temporary virtual patches.
Unit 42 first introduced a one-time assessment capability in April, then began running it against the two cyber-specialized models in August. According to Palo Alto Networks, that assessment uncovered exposures in every customer tested, with 37 percent classified as high or critical severity.
The continuous service represents six months of development work and a $17 million investment, refined across more than 100 customer engagements. When deployed internally, the system identified as many vulnerabilities in three weeks as the organization typically discovers over an entire year.
AI has created an asymmetric advantage for threat actors against organizations trying to defend at human speed. Modern cybersecurity requires machine-speed defense. We're combining Unit 42's offensive testing and threat intelligence expertise with industry-leading AI harnesses and exclusive access to gated capability models to give defenders back the upper hand.
Sam Rubin, Senior Vice President of Unit 42 at Palo Alto Networks
The urgency behind this launch reflects the accelerating threat landscape. Palo Alto Networks reports that attackers using AI have compressed the window between vulnerability discovery and active exploitation by nearly 97 percent in certain scenarios, collapsing what once took weeks into a matter of hours.
Anthropic's Claude Mythos has independently demonstrated its security research capabilities, identifying over 10,000 high-severity vulnerabilities in mainstream software. The model's access has been subject to regulatory constraints: Anthropic temporarily halted Mythos access in June to comply with US export controls, resuming availability on July 1 once those restrictions were lifted.
The service operates on an annual subscription model with pricing that adjusts based on each customer's chosen combination of Anthropic, OpenAI, and open-source models. The offering is available globally.
Source: The Next Web
