The adaptive AI inference security company built to test, defend, and govern enterprise AI deployments — now operating as part of F5.
www.f5.com ↗CalypsoAI was founded in 2018 in Silicon Valley by Neil Serebryany and a team of AI, data science, and machine learning specialists, with a mission to establish the AI Security (AISec) category. The company pioneered LLM security for enterprise and government — offering a test-defend-observe stack that enabled organizations to pressure-test AI applications before deployment, block real-time threats at inference, and maintain full auditability of AI interactions. CalypsoAI built an early customer base in highly regulated sectors, including financial services, insurance, and government agencies, with early validators including Palantir.
In June 2023, CalypsoAI raised $23 million in a Series A-1 round led by Paladin Capital Group, with participation from Lockheed Martin Ventures, Hakluyt Capital, and Expeditions Fund — bringing total funding to $38.2 million. The company expanded its global footprint by opening co-headquarters in Dublin, Ireland, targeting European enterprise and government markets. On September 11, 2025, F5 (NASDAQ: FFIV) announced the acquisition of CalypsoAI for $180 million in cash. The deal closed September 29, 2025 — part of F5's most aggressive M&A year on record — and CalypsoAI's capabilities were rebranded as F5 AI Guardrails and F5 AI Red Team.
CalypsoAI's core technical architecture centered on three capabilities: Red-Team (proactive adversarial testing using swarms of autonomous agents generating 10,000+ new attack prompts monthly), Defend (real-time inference-layer guardrails that detect prompt injection, jailbreaks, data leakage, and policy violations), and Observe (full auditability and traceability of AI interactions for compliance with GDPR and the EU AI Act). The platform was model-agnostic — compatible with GPT-4, Claude, Gemini, and open-source models — and operated across SaaS, on-premises, and hybrid environments. F5 integrated these capabilities into its Application Delivery and Security Platform (ADSP), positioning itself as the first major vendor to embed comprehensive AI-native security directly into application delivery infrastructure.
CalypsoAI's acquisition at $180 million is one of the clearest signals that AI security is graduating from a niche startup category to a core enterprise platform feature. The $180 million price tag — for a company with roughly 50-65 employees and $7.4 million in revenue — reflects F5's conviction that inference-layer security will be as fundamental to AI deployments as WAF protection was to web applications. CalypsoAI's test-defend-observe framework addressed the full AI lifecycle, not just pre-deployment scanning, which made it a structurally differentiated asset. Its IDC recognition in the April 2025 'IDC Innovators: AI Red Team Platforms' report and Gartner's 2022 Cool Vendor designation validated the technology before the acquisition. For CISOs and investors tracking AI security, CalypsoAI's trajectory — from $38 million raised to a $180 million exit in under two years of serious LLM market activity — shows how rapidly the buyer-side demand for AI inference security has crystallized.
As an independent company, CalypsoAI competed against a fragmented field of LLM security startups including Protect AI, Lakera, Robust Intelligence (acquired by Cisco), and Arthur AI. Its differentiation — particularly the proactive red-team capability and inference-layer defense in a single model-agnostic platform — set it apart from point solutions focused solely on guardrails or monitoring. The F5 acquisition resolves CalypsoAI's go-to-market constraint: as a sub-100-person company competing in an enterprise market, distribution was its primary bottleneck. Inside F5, CalypsoAI's technology reaches F5's existing BIG-IP and NGINX customer base — representing a massive installed base of enterprise network infrastructure. The risk is platform consolidation diluting CalypsoAI's identity, but F5's roadmap signals deeper integration rather than sunsetting, and the model-agnostic positioning remains intact post-acquisition.
206 companies across 10 categories — the most comprehensive AI security company tracker.
Browse All Companies →