F5 AI Guardrails

F5 AI Guardrails (the product formerly associated with CalypsoAI, whose domain now serves F5's product page) is a runtime security product from F5 for deployed AI models, apps, and agents. It inspects AI interactions to block prompt injection, jailbreaks, data leakage, and harmful outputs, and logs enforcement actions for audit.

What It Is

F5 AI Guardrails is an AI security and governance layer that enforces policy on model inputs and outputs at runtime. According to F5, it is model-agnostic and applies the same policies to public and private models, as well as to OpenAI, Anthropic, or similarly formatted agents.

How the Guardrails Work

Teams can use prebuilt guardrails or create custom ones through a natural-language interface, tailored by use case, region, and industry. Documented controls include:

• Defense against prompt injection, jailbreaks, and data exfiltration, backed by a continuously updated attack library (F5 states over 10,000 attack patterns are added monthly).
• Semantic data security for standard and custom sensitive-data categories, with controls for PII, PCI, PHI, and EU AI Act alignment.
• Content moderation for toxic, biased, or harmful outputs.
• Agent controls that audit and block unauthorized tool calls and prevent privilege escalation.

Observability and Compliance

Every enforcement action is logged with guardrail attribution so reviewers can see why an interaction was blocked or allowed. Automated auditing templates cover GDPR, HIPAA, and the EU AI Act.

Deployment Options

F5 lists deployment in AWS, Azure, and Google Cloud, in private cloud, and on-premises, including fully air-gapped environments. F5 also publishes the CASI and ARS security leaderboards to compare model risk and resilience, and cites an independent SecureIQLab efficacy report.