Calypso AI

Calypso AI provides an adaptive AI security platform designed to enable organizations to securely manage and govern their artificial intelligence systems, with a particular focus on large language models (LLMs). The company was founded in 2018 by Neil Serebryany, who was inspired to create the company after his work in the US intelligence community at the National Geospatial-Intelligence Agency revealed the significant security vulnerabilities of AI. At the age of 19, while working at the Department of Defense, Serebryany authored a white paper on AI security, and when his recommendations were not implemented, he decided to start CalypsoAI. Having taken university classes at UCLA since age 13, he had already formed connections in the venture capital world who backed his vision.

The company operates on a business-to-business (B2B) software-as-a-service (SaaS) model, generating revenue through subscriptions and licensing agreements for its AI security platform. CalypsoAI targets a diverse client base that includes enterprises in highly regulated industries such as financial services, healthcare, pharmaceuticals, and insurance, as well as government and military agencies. These organizations require robust security to protect sensitive data and ensure compliance as they adopt AI technologies. In September 2025, CalypsoAI was acquired by F5 Networks for $180 million.

CalypsoAI's core product is a comprehensive platform that offers a suite of solutions for testing, validating, and monitoring both internally developed and third-party AI applications. The platform is model-agnostic, meaning it can support various emerging AI models, and provides tools for AI risk management, model validation, and security posture assessment. Key features include modules named 'Red-Team' for simulating attacks, 'Defend' for real-time threat interception, and 'Observe' for comprehensive monitoring and logging. This allows organizations to detect vulnerabilities, understand data biases, prevent data leakage, and defend against adversarial machine learning attacks and prompt injections before deploying models widely. The platform provides real-time visibility into model behavior and customizable policy-based access controls to ensure governance and security across the AI lifecycle.

Keywords: AI security, large language models, LLM security, model risk management, generative AI governance, AI validation, adversarial machine learning, MLSecOps, AI threat detection, enterprise AI security, AI compliance, model testing, data leakage prevention, prompt injection defense, AI risk assessment, responsible AI, secure AI adoption, AI governance platform, cybersecurity for AI, machine learning security