Generative AI adoption is outpacing traditional cybersecurity measures, creating vulnerabilities for enterprises, especially against risks specific to large language models (LLMs). According to Promptfoo’s data, 82% of AI-related security issues in Fortune 500 companies go undetected by standard runtime “guardrails.” Despite investing over 650 hours in manual testing per product release, companies cover only a small fraction, sometimes less than 5%, of critical threats, such as prompt injection, data leakage, and policy evasion.
Founded in 2024, Promptfoo quickly emerged as a leading open-source and enterprise solution for addressing these challenges. More than 125,000 developers worldwide rely on its open-source AI red-teaming and security testing framework. At the same time, over 30 Fortune 500 companies utilise its commercial platform to protect AI systems across various industries, including retail, telecommunications, and finance.
By integrating automated vulnerability detection, reproduction, and remediation into the software development process, Promptfoo helps organisations confidently expand complex AI deployments while protecting critical data and brand reputation.
Experience driving mission-focused innovation
Ian Webster, CEO and co-founder of Promptfoo, brings extensive experience leading AI product teams at scale. Before Promptfoo, he managed LLM product engineering at Discord (200 million users) and worked at Google. He also founded Zenysis (YC W16), a public health analytics platform operating across three continents, and developed Asterank, which Planetary Resources acquired.
His co-founder, Michael D’Angelo, CTO, directed machine learning teams as VP Engineering at SmileID, the world’s largest KYC provider operating in over 20 countries, and co-founded Arthena (YC W17), which was later acquired.
In a conversation with TFN, Ian explains what sparked Promptfoo’s creation: “While leading Discord’s first large-scale LLM features, I witnessed how seemingly benign chatbot prompts morphed into brand-damaging jailbreaks overnight. In one infamous case, a user manipulated the bot into role-playing a grandmother delivering forbidden instructions. At Discord’s scale, a one-in-a-million failure happened hundreds of times daily, exposing a critical gap: the QA and security tools we used for web or mobile apps simply didn’t exist for generative AI.”
Recognising this problem affected all companies integrating LLMs, Ian built an adversarial testing tool that quickly gained traction beyond his team. “Turning that into Promptfoo was a natural step. Our mission is simple: make shipping secure AI as routine as running unit tests,” he adds.
Promptfoo currently employs 11 full-time team members, with a technical staff of three women and six men. The company fosters an inclusive culture focused on technical excellence and rapid innovation in a fast-paced startup environment.
Behind Promptfoo: The community-driven, end-to-end AI security platform
Promptfoo’s platform stands out through its combination of community-led development, sophisticated attack simulation, and seamless enterprise integration. Its open-source version has attracted over 125,000 users, generating valuable feedback that accelerates product improvements aligned with real-world AI security challenges.
A key advantage is Promptfoo’s ability to automatically generate thousands of attack goals and conduct thorough exploit searches for each one. Unlike competitors that rely on manual attack setups and limited testing scopes, Promptfoo utilises research-backed machine learning methods to identify vulnerabilities at scale.
The platform also offers domain-specific attack functions for highly regulated industries, such as finance, healthcare, and insurance, addressing industry-specific compliance and risk concerns. Fully integrated, Promptfoo embeds automated red-team testing within the CI/CD pipeline, provides clear remediation advice, and connects with existing vulnerability management tools to complete the detection-to-resolution process.
This developer-focused approach simplifies the security of complex AI architectures, including Retrieval-Augmented Generation (RAG), agentic systems, and Multi-Channel Processing (MCP).
Ian summarises: “Promptfoo is built on feedback and data from an unprecedented breadth of use cases and architectures, enabling us to focus on the actual security problems enterprises face rather than theoretical ones. Our comprehensive workflow and automation make it possible to ship safe AI at scale.”
Scaling to define the AI security standard
With $23.7 million in total funding, including an $18.4 million Series A led by Insight Partners with participation from Andreessen Horowitz, Promptfoo is poised to accelerate growth. This capital will fund expanded hiring, new feature development, and enhanced enterprise support to meet growing demand from Global 2000 companies prioritising GenAI security.
“Our vision for the next three to five years is clear: become the all-in-one GenAI application security solution. As enterprises increasingly embed AI in core business systems, security can’t be an afterthought. We’re committed to making secure, intelligence-first AI innovation the new normal,” Ian states.
