AI Governance & Security Vendor Wiki

7 min read

CalypsoAI

Researched 2026-06-28; re-verified 2026-07-27. Primary category: ai-runtime-security. Acquired by f5 (closed 2025-09-26). Brand fully retired — calypsoai.com now 301-redirects to F5’s AI Guardrails product page. This is a historical page: you cannot buy CalypsoAI, only F5 AI Guardrails.

One-liner — Formerly an AI runtime-security / “guardrails” pure-play (real-time threat defense + automated red teaming for model traffic); since 2025-09-26 it is the AI-security core of f5’s platform, sold only as F5 AI Guardrails and F5 AI Red Team.

What it does

CalypsoAI built 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 was consistent protection across whatever models an enterprise uses, decoupled from any single model provider.

That technology ships today as F5 AI Guardrails. As of 2026-07-27 the product page advertises adversarial-attack defense (prompt injection, threat research from F5 Labs, model risk scoring via the CASI Model Risk Leaderboard), model-agnostic policy enforcement with natural-language custom policies, prebuilt compliance controls (PII, EU AI Act, PCI, PHI), content moderation, agentic-AI controls that restrict agent actions and tool privileges, and audit-ready logging with visibility into system prompts and tool calls — across public cloud, private cloud, and on-prem. F5 AI Red Team is now a separate SKU in the same line.

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 only 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). There is no standalone CalypsoAI product, contract, or price list any more — procurement runs through F5.

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).

As of 2026-07-27 this is no longer an independent position in the market. The differentiation now belongs to F5: guardrails and red teaming are two SKUs inside ADSP, and F5 has since added surepath-ai’s shadow-AI discovery in front of them. Evaluate the F5 platform story, not a pure-play. The tradeoff shifts accordingly — you gain ADSP integration and a public vendor’s support model, and you lose the independence and pace that made the pure-play attractive. Former nearest neighbors, which remain the real shortlist if you are not an F5 shop: 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 — announced 2025-09-11, closed 2025-09-26. This confirms the seed flag (“acq by F5”). Confidence: high (SEC filing).

On price, two figures circulate and they measure different things:

  • ~$180M — the headline in F5’s 2025-09-11 announcement: “$180 million in purchase consideration financed primarily with cash.” Nearly all press coverage (GeekWire, SDxCentral, Frontier Enterprise) traces to this number.
  • $145.2M cash — what F5 actually recorded at closing, per Note 4 of its FY2026 Q1 Form 10-Q: “On September 26, 2025, the Company closed on a transaction for the acquisition of CalypsoAI Corp. … for $145.2 million in cash.” Allocation: $14.2M net tangible assets, $16.9M developed technology, $114.2M goodwill (total $145.207M, preliminary).

F5 does not reconcile the ~$35M gap, so this page states both with their basis rather than guessing at holdbacks or retention. The GAAP figure is the better number for “what was paid.” F5 also disclosed that CalypsoAI’s revenue and earnings were not material to its operations — a useful reality check on the pure-play’s scale at exit.

Correction (2026-07-27): this page previously recorded the close as 2025-10-08. No primary source supports that date — F5’s completion blog was published 2025-09-29 and the 10-Q states 2025-09-26. Corrected to 2025-09-26.

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. Resolved 2026-07-27: fully retired. calypsoai.com 301-redirects to the F5 AI Guardrails product page, which carries no CalypsoAI branding at all.
  • Exact total funding (sources say “$40M+”; not itemized to a precise figure).
  • What accounts for the ~$35M gap between the announced ~$180M consideration and the $145.2M recorded at close. F5 has not broken this out; unresolved and possibly never disclosed.
  • Whether this page should eventually be merged into f5 rather than kept as a standalone historical entry — it is retained for now because the survey and the M&A map still reference the CalypsoAI name.

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.
  • [2026-07-27] URL audit found calypsoai.com 301-redirecting to https://www.f5.com/products/ai-guardrails. Re-verified against F5’s FY2026 Q1 Form 10-Q and corrected two facts: close date 2025-10-08 → 2025-09-26, and price recorded as $145.2M cash alongside the ~$180M announced consideration. Website repointed to the F5 product page; positioning rewritten as a historical/absorbed entry; brand-retirement open question resolved (fully retired, no CalypsoAI branding remains). Same corrections applied to f5 and the M&A map.

Graph View

Backlinks