The useful question is what changes for users, developers or buyers, and whether the announcement stays industry context or becomes something people can actually use.

AI agents like Claude Cowork are now performing multi-step tasks across enterprise systems, prompting organizations to prioritize security over initial adoption. While Anthropic’s Cowork includes role-based permissions and connector restrictions, these controls only manage access rather than monitor real-time actions within sessions. This leaves gaps where sensitive data could be exposed or automated workflows hijacked by malicious instructions, creating risks for data loss and unauthorized actions. Traditional access controls fall short because AI systems dynamically combine tools, data, and reasoning paths at runtime, making it difficult to predict or govern outcomes after access is granted. To address this, Palo Alto Networks’ Prisma AIRS AI Gateway acts as a centralized control plane, inspecting every AI interaction for sensitive data, enforcing agent identity controls, and providing audit trails for security teams. The gateway also centralizes cost tracking and policy enforcement across all AI tools, enabling consistent governance without disrupting existing workflows or requiring individual tool configurations.

Palo Alto Networks’ solution addresses two critical gaps in AI security: governance visibility and runtime protection. Governance gaps arise when security teams lack insight into data flows through AI tools, such as when a finance analyst uploads unannounced acquisition figures to a model. Runtime risks emerge when automated agents execute actions based on hidden instructions, like a scheduled task modifying files after reading embedded commands in a webpage. These scenarios highlight the need for real-time monitoring and policy enforcement, which traditional access controls cannot provide. The AI Gateway bridges this divide by evaluating every request and tool call, detecting sensitive data like financial projections or contract terms, and ensuring every automated action is traceable to a specific workflow and owner.

The AI Gateway also simplifies enterprise-wide AI governance by consolidating controls into a single console, eliminating the need to configure each AI tool individually. This centralized approach allows CISOs to set spending limits, enforce security policies, and generate procurement reports across all AI tools, including coding assistants, SaaS integrations, and custom agents. For organizations scaling AI adoption, the gateway ensures consistent security and compliance without requiring changes to existing workflows or tools. By rerouting traffic between providers for cost or compliance reasons, it also future-proofs AI investments against evolving vendor landscapes. This scalability is essential as enterprises increasingly rely on multiple AI systems beyond Claude Cowork.

Palo Alto Networks positions its AI Gateway as the foundation for securing the entire AI enterprise, not just individual tools. By deploying Cowork behind the gateway, organizations gain runtime security, data protection, and full visibility across all AI interactions. The same gateway secures other AI tools in the environment under unified policies, reducing operational overhead and ensuring compliance. This centralized control plane addresses the core challenge of governing AI activity at scale, where traditional per-tool security measures fail to keep pace with dynamic, agentic workflows. For enterprises transitioning from pilot projects to full-scale AI adoption, the gateway provides the necessary oversight and control to mitigate risks while maintaining productivity.