You need to ensure consistent threat prevention across all applications. Which approach should you use?

• A. Disable inspection
• B. Use Security Profiles Group
• C. Apply profiles per application manually
• D. Use NAT rules

A global manufacturing organization with 50,000 employees spanning 35 countries designs advanced industrial equipment and owns significant intellectual property. The organization operates in a highly competitive market where protecting trade secrets is critical to maintaining market advantage.
Over the past 18 months, the CISO discovered that employees across the organization have adopted hundreds of GenAI applications to improve productivity. Engineers use AI coding assistants to accelerate product development sales teams use AI tools to generate proposals, and customer service representatives use chatbots to draft responses. While this adoption has driven innovation, it has also created significant security risks.
A security audit reveals sensitive CAD files uploaded to image-generation services, proprietary source code shared with public coding assistants, and confidential customer information used in prompts. The audit identifies over 300 different GenAI applications in use, most of which had not been formally reviewed or approved.
The customer service department has also been developing internal AI applications, including a customer service copilot built on a cloud large language model (LLM) platform, an internal knowledge management assistant, and a code review tool. These internal applications access sensitive databases, customer records and internal APIs - creating additional security concerns about exploitation or misuse.
The organization has a distributed workforce in which 60% of employees work remotely or in hybrid arrangements, accessing corporate resources and AI applications from various locations using managed and unmanaged devices. Existing network security infrastructure lacks AI-specific security capabilities.
Organization leadership wants to enable AI-driven innovation while implementing comprehensive security controls. The CISO has been tasked with developing an organization-wide GenAI governance program that protects sensitive assets without hindering productivity. The program must address both external AI applications employees are using and internal AI applications being developed by IT.
In which two ways would Prisma AIRS secure AI agents deployed across multiple cloud platforms in this scenario? (Choose two.)

• A. By requiring separate product installations for each cloud platform with AWS-specific agents for Bedrock and GCP-specific agents for Vertex AI that cannot share policies.
• B. By offering Network Intercept for infrastructure-level protection across any cloud platform and API Intercept for application-level security embedded directly in agent code.
• C. By providing Network Intercept inline in multicloud network architectures to monitor AI agent traffic, and API Intercept as Security as Code (SaC) to scan prompts and responses before they reach models.
• D. By supporting API Intercept for Multicloud deployments since Network Intercept cannot be deployed in the network architectures of different cloud providers.

You must ensure high availability for critical firewall deployments. What configuration should you implement?

• A. Static routing only
• B. Active/Passive HA
• C. Single firewall
• D. Manual failover

A global organization is modernizing its data center and private cloud infrastructure. The environment consists of:
- A Nutanix AHV cluster hosting critical east-west application workloads
- A VMware ESXi cluster with multi-socket hosts, supporting high-throughput workloads (>10 Gbps)
- A new pair of PA-5450 firewalls to secure the perimeter and handle encrypted traffic inspection at scale
- Strict performance service-level agreements (SLAs) for both north-south and east-west flows, with heavy reliance on TLS 1.3 and IPSec
- A Network Functions Virtualization (NFV) environment on KVM to provide high-performance security services to maximize packet throughput and minimize latency The chief architect is tasked with ensuring that the firewall design avoids hypervisor contention optimizes non-uniform memory access (NUMA) and uses hardware features for encrypted traffic.
VM-Series on Nutanix AHV - Resource Allocation
- Because the Nutanix cluster is already heavily used, the architect's main concern is preventing performance degradation of the virtual firewall. Thin provisioning or ballooning could introduce latency and unpredictability which is unacceptable for a security-sensitive workload.
VM-Series on VMware ESXi - NUMA and vCPU Placement
- In the VMware ESXi environment, the architect is deploying VM-Series for workloads pushing >10 Gbps. Assigning vCPUs across NUMA nodes or oversubscribing cores would create latency due to cross-socket memory access and scheduling delays. Similarly, dedicating logical hypethreads does not provide the deterministic data plane performance required.
Operational Integration and High Availability
- With performance guaranteed by correct hypervisor and hardware provisioning, the architect also considers high availability (HA). VM-Series pairs are deployed in active/passive HA across Nutanix and VMware clusters, while PA-5450s form the data center's north-south secure perimeter deployment. This ensures resilience without introducing unnecessary east-west inspection bottlenecks.
- The recommendation must be a scalable, high-performance firewall deployment aligned with enterprise SLAs and the CISO's encrypted traffic concerns.
To optimize throughput and minimize latency, what is recommended to configure the vCPUs and NUMA for this deployment?

• A. Ensure that all vCPUs assigned to the VM's data plane reside on a single physical NUMA node
• B. Configure the number of vCPUs to be greater than the number of physical cores on the host in order to use the ESXi scheduler
• C. Enable hyperthreading on the physical host and assign all logical cores from a single physical core to the VM-Series
• D. Assign vCPUs from multiple NUMA nodes to allow the VM to access more memory