Global Bank Maintains AI Security at Scale with F5

A global investment bank needed to secure a quickly growing enterprise AI platform without sacrificing threat detection accuracy as workloads scaled. F5 AI Guardrails gave the bank’s cyber controls team a security layer built to maintain peak protection while supporting a rapidly expanding portfolio of AI applications.

Business Challenge

Global investment banks operate some of the most demanding technology environments in financial services, processing enormous volumes of data and transactions around the clock while remaining accountable to regulators across multiple jurisdictions. When one of the world’s largest global investment banks began rolling out a generative AI platform across the enterprise, they needed security that could keep pace without becoming a bottleneck.

The bank’s cyber controls organization owned the mandate to secure every AI application before it reached production. Live applications were already processing roughly 100,000 AI prompts a day, with an enterprise-wide ambition to extend that same coverage to hundreds of applications. The central technical challenge became clear: Any security layer sitting in the path of that traffic had to maintain peak threat detection accuracy, measured by its F1 score, even as workloads escalated. A security solution that added latency or delivered lower detection accuracy under heavy loads would become unusable or increase the risks for data leakage, prompt injection, and other AI-specific threats.

With performance and threat detection accuracy as the core technical challenges, the bank also needed a solution built to satisfy strict global regulatory expectations, including auditable logs of AI traffic, clear policy enforcement records, and forensic investigation capability. Given the bank’s plans to bring agentic AI infrastructure online, agility is essential to its security strategy as use cases extend beyond conversational AI toward more autonomous, workflow-driving systems.

The bank ran a rigorous, multi-phase evaluation before moving to production, testing both cloud and on-premises deployment models to meet its performance, security, and data control requirements across deployment environments and at enterprise scale.

Solutions

The bank selected F5 AI Guardrails, self-hosted within the enterprise cloud environment rather than as a SaaS service, giving the cyber controls team full command over where AI traffic is processed and how data is handled. The deployment will run inline, sitting directly in the data path across AI applications, with GPU-backed infrastructure and Kubernetes-based orchestration to handle load balancing and autoscaling as usage grows.

Results

Protect data across every AI interaction

AI Guardrails inspects every AI interaction inline across the bank’s live production applications and monitors full AI conversation sessions in real time. Prompt injection protection applies to every app and PII protection is layered on top for highly sensitive use cases. Initial deployment in audit mode demonstrated meaningful improvements in the accuracy of threat detection while meeting the bank’s core requirement of maintaining peak F1 accuracy even as workloads expand. Continued investment and collaboration with F5 on GPU capacity and orchestration have brought response times down further as traffic volumes grow and the bank scales toward deployment of hundreds of secure, high-performance AI-enabled applications without compromising the depth of inspection each AI interaction receives.

Maintain peak detection accuracy at scale

As the bank scales their AI platform from a handful of applications to hundreds, centralizing security with AI Guardrails enables them to proceed with confidence. This AI security model gives the bank’s cyber controls team a consistent way to bring new AI applications online safely rather than building bespoke protections for each one.

Support global regulatory compliance

Every interaction is logged, giving the security team real-time threat visibility and a complete, auditable record of AI traffic across the organization. That includes policy enforcement records and forensic investigation capability that provides the security and compliance teams with the evidence required to meet regulatory expectations across jurisdictional boundaries.

Gain a foundation for platform-based security

As the AI platform continues to grow, the bank is considering F5 AI Usage Control, a solution designed to help surface all shadow AI usage across the organization. AI Usage Control represents another component of the broader F5 AI Security Platform, which offers to unify AI Usage Control, AI Guardrails, and F5 AI Red Team on a single control plane through F5 AI Gateway. This platform-centric approach would pave a path to the strongest AI security posture available.