The correct answer is B . In Zero Trust architecture, content stored in Software as a Service (SaaS), Platform as a Service (PaaS), or Infrastructure as a Service (IaaS) environments should not be assumed safe simply because it resides in a cloud platform. Zscaler’s security model emphasizes that trust must be established through inspection and policy , not by location alone. The TLS/SSL inspection architecture shows that inline inspection is necessary to evaluate content moving through encrypted sessions, while Zscaler’s broader data protection model also includes out-of-band assessment for content already stored in cloud services.

This aligns with the Zero Trust principle that applications and content can exist anywhere, but they are not automatically trustworthy because of where they are hosted. Cloud providers secure the platform, but they do not guarantee that every uploaded file, shared object, or stored dataset is safe, compliant, or free from malware or data exposure risk. At the same time, saying content should never be trusted is too absolute; Zero Trust is about verification , not blanket denial. Therefore, the most accurate answer is that cloud-stored content should be treated as risky until inspected , whether inline during transfer or out of band while at rest.

The correct answer is A . In Zero Trust architecture, content inspection is most effective when it happens inline at the policy enforcement point and as close to the initiator as possible . This improves both security and user experience. From a security standpoint, inspecting traffic early allows the platform to identify malware, risky content, command-and-control behavior, and sensitive data movement before the traffic continues deeper into the environment or reaches the destination. From a performance standpoint, enforcing policy at the nearest edge reduces unnecessary backhaul and helps maintain a more efficient path.

This aligns with modern cloud-delivered Zero Trust design, where users connect to the nearest enforcement point rather than being forced through a central data center stack. A one-armed concentrator model is a legacy deployment concept and is less effective for distributed users and applications. Inspecting data only after it has been copied to disk is too late for inline protection, and an ISP backbone is not the enterprise’s policy enforcement location. Therefore, the best answer is that content should be assessed at the enforcement point closest to the initiator , such as the nearest service edge.