OpenAI Acquires Promptfoo to Enhance Security for Its Enterprise AI Agents

San Francisco: OpenAI has agreed to acquire Promptfoo, an AI security startup founded in 2024 by Ian Webster and Michael D’Angelo, and will integrate its technology directly into OpenAI Frontier, the company’s enterprise platform for building and operating AI agents.

Financial terms were not disclosed. Promptfoo had raised $23 million in total funding, including an $18.4 million Series A in July 2025 led by Insight Partners and Andreessen Horowitz. PitchBook data places its last valuation at approximately $119 million. The startup’s 23-person team will continue building inside Frontier after the deal closes.

Webster conceived the idea while running Discord’s engineering team, shipping AI products to 200 million users with no reliable way to test whether those systems would hold up under adversarial pressure. The platform he and D’Angelo built became one of the most widely adopted AI security tools in enterprise software.

Promptfoo’s open-source CLI and library are now used by more than 25% of Fortune 500 companies, with over 350,000 developers having pulled it into their workflows and 130,000 active monthly users. Its core function is adversarial evaluation, systematically probing AI systems for prompt injections, jailbreak vulnerabilities, data leaks, tool misuse, and out-of-policy agent behaviors before those systems reach production.

That capability is precisely what OpenAI needs as enterprises begin deploying AI agents, what the company calls “AI coworkers” into live business operations. OpenAI launched Frontier on February 5, 2026, with early customers including Uber, State Farm, Intuit, and Thermo Fisher Scientific.

As those deployments deepen, the attack surface expands. Agents connected to real data, internal tools, and external APIs introduce security risks that static model testing cannot catch. Promptfoo’s integration into Frontier will make automated red-teaming, agentic workflow evaluation, and compliance reporting native features of the platform rather than third-party add-ons.

“As AI agents become more connected to real data and systems, securing and validating them is more challenging and important than ever,” said Webster. “Joining OpenAI lets us accelerate this work, bringing stronger security, safety, and governance capabilities to the teams building real-world AI systems.”

OpenAI has committed to maintaining Promptfoo’s open-source project alongside the proprietary enterprise features it will build into Frontier. The acquisition follows a clear pattern of OpenAI using targeted M&A to close product gaps at speed earlier this year the company acquired healthcare tech startup Torch for approximately $100 million, and previously bought Software Applications, maker of the Sky AI interface for Mac.

It also arrives as rival Anthropic launched Claude Code Security in February 2026, targeting the same vulnerability scanning space, signaling that enterprise AI security is fast becoming a competitive battleground among frontier labs.