OpenAI Expands Enterprise AI Push With Promptfoo Security Startup Deal

OpenAI disclosed this week that it plans to acquire AI security startup Promptfoo, folding the San Mateo, California-based company’s testing and red-teaming tools into its enterprise agent platform, Frontier.

Founded in 2024 by Ian Webster and Michael D’Angelo, Promptfoo built a reputation among developers for tools that stress-test large language models and agent systems. Its software focuses on identifying vulnerabilities such as prompt injection attacks, jailbreak attempts, data leaks and misuse of external tools — issues that can surface when AI systems connect to real-world data and applications.

Once the deal closes, the roughly 23-person Promptfoo team will join OpenAI. The companies said Promptfoo’s open-source tools will continue to be maintained and supported, meaning existing developers and organizations using the platform should see no disruption.

Promptfoo’s technology is widely used to simulate adversarial attacks against AI systems. Its platform includes automated red-teaming, code scanning, vulnerability detection and remediation tools that integrate directly into development pipelines such as Github, Gitlab and Jenkins.

The company claims adoption across a growing developer community, including more than 350,000 developers overall and more than 130,000 active monthly users. Promptfoo also reports that 127 Fortune 500 companies — including retailers, telecom providers and healthcare firms — rely on its tools for AI testing and evaluation.

Srinivas Narayanan, OpenAI’s CTO of B2B applications, said the acquisition will help strengthen Frontier’s built-in evaluation and security capabilities.

“Promptfoo brings deep engineering expertise in evaluating, securing, and testing AI systems at enterprise scale,” Narayanan said in a statement. “We’re excited to bring these capabilities directly into Frontier.”

Promptfoo CEO Ian Webster said integrating with OpenAI could accelerate efforts to secure increasingly complex AI deployments. As agents gain access to enterprise data and operational systems, validating how they behave becomes far more challenging.

“As AI agents become more connected to real data and systems, securing and validating them is more challenging and important than ever,” Webster said. “Joining OpenAI lets us accelerate this work.”

With that shift comes heightened concern over security risks, from data leakage to agents taking unintended actions inside corporate systems. Analysts say automated red-teaming and evaluation tools are becoming essential infrastructure for enterprise AI deployments.

For OpenAI, the acquisition adds a security layer to Frontier at a moment when enterprise adoption of AI agents is accelerating — and when companies deploying those agents want fewer surprises and more guardrails.