- Category
- AI runtime security
- Website
- f5.com
- Last verified
- November 1, 2026
What is F5 AI Guardrails?
F5 AI Guardrails and F5 AI Red Team are F5's AI security products, built on technology from CalypsoAI, a Dublin-based AI security company. F5 announced it had completed the acquisition in September 2025, and press reports put the price at about $180 million. F5 describes AI Guardrails as runtime security for AI models and agents, and is integrating both products into its Application Delivery and Security Platform. The old calypsoai.com address now redirects to F5's product page.
Key capabilities
- Custom guardrails written in natural language.
- Prompt injection and jailbreak protection, backed by a continuously updated attack library.
- Agent controls that audit and block unauthorized tool calls and actions, and log prompts, reasoning and tool calls across single- or multi-agent systems.
- Data-leak prevention and content moderation, with on-premises and air-gapped deployment options.
- AI Red Team, which uses swarms of autonomous agents to simulate attacks and can turn findings into guardrails.
Why it matters for humanoid robot security
Editorial analysis: F5's products are aimed at enterprise AI, not robots. The relevant feature for humanoids is the ability to audit and block an agent's tool calls at runtime. When a robot's planner is an LLM agent, its tool calls are its actions. A policy layer that can refuse an unauthorized action, and log the reasoning that led to it, is the kind of control safety and security teams will want between the model and the motors. Air-gapped deployment also matters for factories that keep robots off the internet.
Where it fits in a humanoid robot security program
Analysis: F5 AI Guardrails sits between the AI agent and the tools it calls, which in a humanoid means between the planner and the robot's skill library. Its main stage is operations, as a runtime filter and audit trail, with AI Red Team used during testing to find weaknesses before deployment. Teams would need to confirm latency and on-robot or on-site deployment fit for real-time use.
Who it is for
- Enterprises already using F5 that are adding LLM agents to robots or automation
- Robot software teams that need an audit log of agent decisions
- Security teams in air-gapped industrial sites
Questions to ask F5 AI Guardrails
- Can guardrails run on-site or on an edge server with low enough latency for robot control loops?
- How do you block or modify a specific tool call without stopping the whole agent?
- Can red team findings be converted into enforceable guardrails automatically?
- What logs do you keep for incident investigation after an unsafe action?
Similar listings
- Cisco AI Defense: AI security
- Lakera (Check Point): AI and LLM security
- Zenity: AI agent security
Related on Humanoid Threats
- Prompt injection against robot VLA models
- Can LLM-controlled robots be jailbroken?
- Browse the full Humanoid Robot Security Directory
Sources
- F5 AI Guardrails product page
- F5: F5 completes acquisition of CalypsoAI
- Frontier Enterprise: F5 wraps up US$180M acquisition of CalypsoAI
This listing is editorial and free; inclusion is not an endorsement. Work at F5 AI Guardrails? Claim or update this listing.
Get the weekly Humanoid Threats Brief → Subscribe


