SAN FRANCISCO – Security specialist Wiz (now part of Google Cloud) on Monday announced the AI Application Protection Platform (AI-APP), a solution designed to secure the increasingly complex web of models, AI agents, and data that define artificial intelligence (AI)-native development.

The move, announced at RSAC 2026, marks a significant evolution in the Cloud-Native Application Protection Platform (CNAPP) market. While traditional security tools often analyze infrastructure, identities, and applications in silos, Wiz’s new platform treats AI as a dynamic, interconnected system.

According to Wiz, the “blind spots” created by rapid AI adoption are a primary concern for CISOs. Organizations are currently juggling managed services like AWS Bedrock, SaaS ecosystems like Microsoft Copilot, and custom self-hosted frameworks.

The AI-APP addresses this by building a comprehensive inventory across all environments. A central feature, the Wiz Workload Explainer, leverages AI to scan and translate custom implementations into clear components—identifying agents and data flows that traditional deterministic scanning often misses.

The core philosophy of the new platform is that AI risk is rarely the result of a single vulnerability. Instead, risk emerges when benign conditions across different layers overlap.

“AI risk is not defined by a single issue, but by how multiple conditions come together,” the company said in a blog post. By correlating signals across the application stack, Wiz can map real, exploitable attack paths. For example, the platform can identify if an AI agent has the permissions to execute code or modify infrastructure—actions that, if manipulated, could lead to catastrophic breaches.

To help teams prioritize these threats, Wiz maps identified risks against the OWASP Top 10 for LLM Applications, ensuring security teams focus on vulnerabilities like prompt injection and data leakage within a recognized compliance framework.

The platform introduces a triple-layer threat detection strategy: Model Activity, for monitoring inputs/outputs and prompt behavior; Workload Execution, to track agent activity and tool usage; and Cloud Layer, which observes API calls and identity changes.

To bolster this defense, Wiz has integrated with key partners in the Wiz Integration Network (WIN). Collaborations with Cloudflare, TrojAI, and Pillar Security allow the platform to ingest external red-teaming findings and endpoint security data, providing a “single pane of glass” for AI security.

Separately, Wiz unveiled Red Agent, an AI-powered attacker that acts as a sophisticated security researcher, but with AI speed and scale.

The Red Agent joins Wiz’s Green and Blue Agents to power agentic workflows that allow teams to operate how AI operates within their environment.