AI Security Platform Promptfoo Raises $18.4 Million Series A to Build AI Security Stack

Promptfoo, a full-stack AI Security Platform, led by Ian Webster(CEO & Co-Founder), Michael D'Angelo(CTO & Co-Founder), and others, raised $18.4 million Series A led by Insight Partners, with participation from Andreessen Horowitz. Ganesh Bell, Managing Director at Insight Partners, joins the board.

AI security is now the biggest barrier to deploying generative AI in enterprises. New architectures like agents, RAGs, and MCP have unlocked powerful capabilities—but also expanded the attack surface. Systems now face risks like prompt injection, data leaks, and insecure tool use across text, voice, and image inputs. Traditional security tools can’t handle this complexity, and manual testing falls short.

The internet is shifting toward agent-based systems powered by MCP servers, with agents that collaborate, run code, and access sensitive data. Enterprises are embedding proprietary information into agents and RAGs, introducing new security risks. Key trends include: deeper support for multi-agent workflows with safety checks; full observability of MCP tool calls; built-in security for audio-first interfaces; and a shift toward test-driven prompt development. Regulation is catching up, with the EU AI Act and U.S. frameworks like OWASP LLM Top 10 and NIST setting new standards.

Red teaming is becoming essential in the AI development cycle, evolving from a specialized task to a standard practice alongside product and quality evaluations. As agents grow more capable, continuous adversarial testing will be as routine as unit testing in software. The goal is to make every model, prompt, or tool change automatically tested, red-teamed, and approved with a signed safety report—integrating security directly into development. This shift helps teams catch vulnerabilities early, build security expertise, accelerate deployment, and meet compliance with ease.