Promptfoo is an AI security company that provides an open-source framework and enterprise platform for testing, evaluating, and securing large language model (LLM) applications. Founded in 2024 and headquartered in San Mateo, California, the company was acquired by OpenAI on March 16, 2026, while remaining open source. The company's core product enables developers and security teams to perform adversarial testing and automated red teaming of AI applications. Rather than relying on static lists of harmful prompts, Promptfoo dynamically generates attacks based on the specific context of the target application, covering risks such as prompt injection, jailbreaks, data leakage, context poisoning, tool misuse, and use-case hijacking. Promptfoo's product suite includes: - An open-source evaluation and red teaming framework used by over 350,000 developers - An enterprise platform with collaboration features for development and security teams - ModelAudit, a static security scanner for ML model files (MIT-licensed) - An LLM-focused code scanner for identifying LLM-related vulnerabilities in application source code - An MCP (Model Context Protocol) Proxy for managing security risks in MCP server deployments The platform is designed to address security challenges specific to modern AI architectures including autonomous agents, RAG (Retrieval-Augmented Generation) pipelines, multimodal systems, and MCP integrations. It targets both individual developers and large enterprises, with adoption reported across more than 25% of Fortune 500 companies. Promptfoo is SOC 2 Type II compliant and ISO 27001 certified. The company raised a $5M seed round from Andreessen Horowitz and an $18.4M Series A led by Insight Partners prior to its acquisition by OpenAI.