OpenAI Buys Promptfoo to Boost Frontier Security

OpenAI has acquired Promptfoo, a tooling startup known for helping developers test and secure prompts used with large language models. Promptfoo counts more than 125,000 developers and over 30 Fortune 500 companies among its users, and its focus on prompt validation and exploit detection makes it a natural fit for OpenAI’s push into application security.

The technology will be integrated into Frontier, OpenAI’s enterprise agent platform launched last month. Frontier is designed to let businesses orchestrate AI agents across workflows, and bringing Promptfoo’s capabilities into that environment aims to reduce risk from prompt‑level vulnerabilities and unexpected model behaviors.

For teams running production agents, prompt safety is becoming as essential as code testing. Promptfoo provides automated checks, regression tests and attack simulations that flag risky prompt patterns or outputs. Adding this to Frontier could help enterprise customers more confidently deploy autonomous AI across sensitive tasks.

OpenAI has increasingly been moving beyond model research into tooling and operational security. Recent announcements around Codex Security and enterprise features paint a picture of a company prioritizing the reliability and safety of deployed AI systems. Promptfoo’s acquisition appears to be a step toward making prompt governance a built‑in part of the agent stack.

Developers who’ve relied on Promptfoo for validation may see the integration accelerate feature availability inside Frontier, while enterprises evaluating agent platforms will likely weigh the added safety assurances as a differentiator. For the broader AI ecosystem, this deal underscores how prompt engineering and application‑level security are maturing into core enterprise concerns.