F5 AI Guardrails vs Prediction Guard

F5 AI Guardrails is F5's runtime security product for AI models, applications and agents. F5 sells it alongside F5 AI Red Team, its automated adversarial testing product, and presents the two as a pair: findings from Red Team can be turned into guardrail policies. Both come from CalypsoAI, the AI security company F5 announced it would acquire on September 11, 2025 for $180 million in purchase consideration, mostly cash. The deal closed on September 26, 2025. F5 introduced both products when it announced the deal had closed, and made them generally available in January 2026 as part of its Application Delivery and Security Platform. F5, Inc. was incorporated in 1996 and is based in Seattle; this profile lists the company's founding year and headquarters, as the directory does for other big-company products. F5 AI Guardrails blocks prompt injection and jailbreaks, detects and prevents leakage of personal data and custom data categories, filters toxic or harmful output, and audits or blocks unauthorized tool calls and actions by agents. Policies can be written in natural language. Every enforcement action is logged with attribution, and logs can be exported to third-party SIEMs. It works with any model and runs in AWS, Azure or Google Cloud, in private clouds, or on premises, including fully air-gapped environments. F5 AI Red Team sends a swarm of attacker agents to run static, prompt-based and multi-turn attacks, including indirect prompt injection and data-leakage tests on RAG applications. F5 says its attack library grows by more than 10,000 patterns a month. Red Team runs in the major public clouds or on premises as a certified Red Hat OpenShift operator. A feature called F5 AI Remediate drafts guardrails from Red Team findings, and a person approves each one before it goes live. **Gateway and platform** On June 22, 2026 F5 launched the F5 AI Security Platform, which groups AI governance, AI discovery, testing with AI Red Team and runtime protection with AI Guardrails, plus an audit trail across AI interactions. At the same time it announced the acquisition of SurePath AI, which adds network-based discovery of AI use, including shadow AI. F5's AI Gateway now combines a model gateway, an MCP gateway (with an MCP server registry and per-tool allow and deny controls) and AI Guardrails under one control plane. **Compliance coverage** F5 says Guardrails has automated auditing templates for GDPR, HIPAA and the EU AI Act, out-of-the-box policy controls aligned to PCI, PHI and the EU AI Act's prohibited-risk categories, and audit-ready logs. It also says Red Team findings can support compliance programs aligned with the NIST AI RMF and the EU AI Act, and it publishes an infographic aligning its products with the OWASP Top 10 for LLMs. We found no article-by-article mapping to the EU AI Act. **Who it's for** Large enterprises, including regulated financial services and healthcare organizations, that need AI guardrails they can host themselves or run air-gapped. F5 says both products are deployed at Fortune 500 companies. It also suits F5 customers who want AI controls on the same platform as their application delivery and API security.

Prediction Guard sells a self-hosted "sovereign AI control plane" for regulated organizations. It sits between applications and the models, agents and MCP tools they call, and applies access, safety and policy controls to each request inside the customer's own infrastructure. The company was founded in 2023 in Lafayette, Indiana; CEO Daniel Whitenack is a co-founder. It raised a $3.7 million seed round led by Sovereign Capital in May 2025, and Noblis Ventures made a strategic investment of undisclosed size in July 2025. At the time of the seed round it had eight employees and said it had 18 customers in health care, defense, transportation and manufacturing. The platform registers each agent with an identity and role and limits it to the models and tools that role needs. Runtime controls include PII masking, prompt injection detection, toxicity filtering, input sanitization, MCP tool scoping and human-in-the-loop approval; requests that fail a policy are blocked and logged. It produces an exportable AI bill of materials (CycloneDX) of models, guardrails and MCP servers, runs model security scans and streams security events to SIEM tools such as Splunk, Datadog and Grafana. Applications call it through OpenAI- and Anthropic-compatible APIs. It deploys into a customer VPC on AWS, Azure or Google Cloud, on Kubernetes on premises, or air-gapped. Prediction Guard says its control overhead is under 200 ms per request. **AI compliance coverage** Prediction Guard says its policies align with the NIST AI RMF, NIST AI 600-1, the OWASP Top 10 for LLM and agentic applications, OMB M-26-04 and AIUC-1, with one-click baselines and an exportable "proof of alignment" for the NIST AI RMF. A company blog post (August 2026) adds the EU AI Act and ISO/IEC 42001. The same post labels its OWASP mapping "self-published" and not independently verified, and notes that the audit log records enforcement events rather than full replayable traces. We found no published control-by-control mapping. The homepage shows badges for a SOC 2 Type 2 report by Sensiba and HIPAA compliance; there is no public trust center. **Who it's for** Regulated enterprises and public sector bodies that want AI guardrails and agent controls to run inside their own network, including disconnected environments. It is sold to government through Carahsoft. Its homepage quotes executives at Noblis (also an investor), SimWerx and iBase-t.