AI Governance & Security Vendor Wiki

3 min read

CalypsoAI

Researched 2026-06-28. Primary category: ai-runtime-security. Acquired by f5 (completed 2025-10-08).

One-liner — An AI runtime-security / “guardrails” pure-play (real-time threat defense + automated red teaming for model traffic), now the AI-security core of f5’s platform.

What it does

CalypsoAI builds model-agnostic security for generative and agentic AI: an inline guardrails layer that screens prompts and responses in real time (prompt-injection defense, data-loss prevention, content/abuse controls) plus red teaming at scale that continuously attacks your AI systems to surface vulnerabilities. The pitch is consistent protection across whatever models an enterprise uses, decoupled from any single model provider.

Where it sits in the stack

Primary ai-runtime-security (the “AI firewall” inline control) with a secondary ai-governance-platform tag (its reporting/red-team posture supports AI governance). Layer: model-prompt. Lethal-trifecta role: addresses all three legs at the inference boundary — untrusted input (prompt injection), sensitive data (DLP on prompts/responses), and egress (policy on what model traffic is allowed). Sits at the green↔red boundary.

Deployment & architecture

SaaS and self-hosted; API-driven and deployable inline in front of model endpoints (model-agnostic across proprietary + open-source LLMs). Post-acquisition it is delivered as F5 AI Guardrails (runtime protection) and F5 AI Red Team (autonomous attack simulation, advertised >10,000 new attack patterns/month), integrated into the F5 Application Delivery and Security Platform (ADSP).

Positioning & differentiators

Known for pairing inline guardrails with continuous automated red teaming in one product — defense + offense — and for model-agnostic coverage. It was a Top-Two Finalist in the 2025 RSAC Innovation Sandbox and named to Fast Company’s Most Innovative Companies in AI (2025) (vendor/press accolades, noted as such). Nearest neighbors: lakera, prompt-security, witnessai, aim-security, hiddenlayer, prisma-airs, enkrypt-ai, splxai.

Ownership, funding & M&A

Founded 2018; principal operations in Dublin, Ireland (with US roots/Silicon Valley presence). Pre-acquisition funding reported at over $40M (Series A $13M led by Paladin Capital Group in 2020; a $23M round announced 2023-06-27; investors incl. Paladin, Lockheed Martin Ventures, Hakluyt Capital). Acquired by f5 for ~$180M — announced 2025-09-11, completed 2025-10-08. This confirms the seed flag (“acq by F5”). Verified against F5’s press release/blog and CalypsoAI/Paladin sources. Confidence: high.

CTO / hedge-fund lens

Day-1 control type (runtime AI firewall) but, as a now-absorbed product, you’d evaluate it as F5 AI Guardrails rather than standalone CalypsoAI. Relevant if you’re an F5 shop or want guardrails + red-teaming bundled. The red-team capability has mild SR 11-7 / model-validation value (evidence that AI systems were adversarially tested). Best fit: enterprises standardizing on F5; smaller funds would weigh it against lighter SaaS guardrails.

Competitors / alternatives

lakera, prompt-security, witnessai, aim-security, hiddenlayer, prisma-airs, cisco-ai-defense, enkrypt-ai, splxai, trojai.

Open questions / to verify

  • Whether the CalypsoAI brand survives or is fully retired into F5 product naming.
  • Exact total funding (sources say “$40M+”; not itemized to a precise figure).

Sources

History

  • [2026-06-28] Stub created from seed registry.
  • [2026-06-28] Researched; confirmed acquired by f5 ($180M; announced 2025-09-11, completed 2025-10-08). Seed flag verified. Filled founding (2018), Dublin HQ, funding ($40M+), product (now F5 AI Guardrails / AI Red Team). Confidence raised to high.

Graph View

Backlinks