Wiz Red Agent, our AI-powered attacker, recently launched in Public Preview to all Wiz ASM Advanced customers, marking a major milestone in our mission to help organizations stay one step ahead of attackers. Our goal is to continuously uncover complex exploitable risks across the modern attack surface at a speed and scale that was previously impossible.
The Red Agent leverages the latest frontier models, including the Anthropic Claude Opus series, to reason like a human attacker- but at AI speed and scale.
Working with Anthropic: The Power of the Frontier Opus models
We are excited to leverage Anthropic’s frontier Opus models, among other models, to power the Wiz Red Agent. With our recent public preview, we have deployed these models at a large scale, scanning over 150,000 production web applications and APIs every week. Integrating its reasoning engine directly into the Red Agent to protect our customers’ most sensitive environments.
By leveraging the Opus model series, the Red Agent analyzes application logic, chains multiple steps together, and dynamically adapts attack patterns based on real-time server responses. This enables it to find the "un-findable" logic flaws that traditional scanning structurally cannot see and protect our customers from emerging risks that could only be discovered manually before the rapid advancements of frontier models.
The Impact: Thousands of High and Critical Exploitable Risks Uncovered across the World’s Largest Organizations
Leveraging the frontier Opus model series and specifically since Claude Opus 4.6 was released, has allowed us to deliver amazing results for our customers. We have worked with hundreds of customers- including some of the world’s largest Fortune 100 security teams- to put Red Agent to the test. Even in these highly hardened environments, Red Agent consistently discovered exploitable risks that teams didn't know existed.
Today, as we scale across hundreds of enterprise environments, we are seeing the impact of frontier intelligence in production:
150,000+ Assets Scanned Weekly: Red Agent runs continuously, analyzing everything from APIs to web apps at a volume manual testing could never reach.
115B+ Tokens Processed Every Week: We are using over a hundred billion tokens on a weekly basis dedicated to finding exploitable risks on publicly facing production environments.
3,000+ High and Critical Risks Every Week: We have identified thousands of validated, multi-step attack chains in production environments.
0 False Positives: After an extensive testing period, the Red Agent has reached a state where it produces 0 false positives, while accompanying every finding with concrete proof of exploitability, impact and business context.
Solving Web Security with Wiz Context + frontier models
We believe traditional web security is a solved problem- by combining deep cloud context from the Wiz Security Graph with the reasoning capabilities of frontier models like Claude Opus 4.6 and 4.7, Red Agent acts like the world’s best human attacker - running on your side as your friendly hacker. Red Agent runs continuously across your entire environment- it doesn't rely on periodic manual testing, but it scales to thousands of APIs and web apps instantly. Frontier AI models like the Opus model series empower it to find complex, logic-driven exploitable risks that would be impossible to find otherwise. And by leveraging context from the Wiz Security Graph, customers identify the true impact of such risks and prioritize, assign, and remove the risk fast.
Get Started with the Red Agent now
Experience what happens when the world’s best attacker joins your team. Red Agent is available now in Public Preview for Wiz customers in the Preview Hub, with many new exciting capabilities on the horizon. Interested in seeing the Red Agent in action? book a live demo with our team.
