OpenAI Acquires Promptfoo to Enhance AI Security for Enterprises
OpenAI has announced its acquisition of Promptfoo, a security platform that specializes in evaluating and testing AI systems in production. The technology will be integrated directly into OpenAI Frontier, the company’s enterprise AI agent platform launched in February 2026. The financial terms of the deal have not been disclosed. For Chief Information Officers (CIOs) and Chief Information Security Officers (CISOs) involved in AI projects, this acquisition signifies a pivotal shift from ad-hoc testing to integrated, native security within the development cycle.
The deployment of AI agents in enterprise environments presents organizations with specific risks not typically found in traditional software applications. Unlike deterministic systems, an AI agent connected to tools, databases, or business workflows can exhibit unpredictable behaviors, execute unauthorized calls, or leak sensitive information. Promptfoo addresses these risks by providing a structured framework to assess large language models (LLMs) and agents in real-world conditions through an open-source command-line interface (CLI) and a dedicated commercial suite for security teams and developers.
Promptfoo’s platform is built on a systematic evaluation architecture that allows AI systems to be tested against structured prompts, adversarial scenarios, and real-use cases. Founded in 2024, the startup is currently utilized by over 25% of Fortune 500 companies. It targets four critical risk categories for organizations deploying agents: prompt injections, jailbreak attempts, data leaks, and out-of-policy tool executions. These attack vectors have become priorities for cybersecurity teams as AI agents increasingly access file systems, APIs, and internal knowledge bases.
Ian Webster, CEO of Promptfoo, stated in the announcement that the startup was created to provide developers with a tangible means to secure AI systems, and its integration with OpenAI will significantly scale this effort. The open-source tool, which covers comparative assessments of models—including ChatGPT, Anthropic’s Claude, and Google’s Gemini—will remain available under its current license. OpenAI is committed to maintaining this community project, according to an official release on March 9, 2026.
Integrating Security and Governance into Development Pipelines
OpenAI Frontier, launched in February 2026, serves as the foundation for the company’s enterprise AI agent offerings, referred to as “AI coworkers.” The system already includes permission mechanisms, shared context, and traceability. The acquisition of Promptfoo adds three operational capabilities: automated security testing and red teaming, integration into development workflows to identify and mitigate risks prior to deployment, and governance tools that meet the increasing compliance, risk, and auditing requirements faced by organizations.
Srinivas Narayanan, CTO of B2B applications at OpenAI, emphasized that Promptfoo brings its expertise in evaluating and securing AI systems at an enterprise scale, with the goal of embedding these capabilities directly into Frontier. For businesses, this means that agents deployed through the platform will undergo standardized validation processes before going live, reducing reliance on manual audits or third-party tools.
Valued at $86 Million with a Lean Team
Promptfoo has raised a total of $22.68 million, including an $18.4 million Series A round in July 2025 led by Insight Partners with participation from Andreessen Horowitz. At that time, its post-money valuation was reported at $85.5 million, according to PitchBook. With a small team of 11 employees, the data highlights the disparity between the startup’s size and its operational footprint: it is present in the AI security environments of one in four Fortune 500 companies, indicating strong adoption driven by the quality of the open-source tool rather than traditional sales efforts.
This acquisition is part of a series of strategic purchases by OpenAI, including Neptune in December 2025 for model governance, Statsig in September 2025 for $1.1 billion to enhance experimentation capabilities, and health startup Torch in January 2026 for approximately $60 million. The acquisition of Promptfoo signals a clear focus on agent security, a segment where competition among AI labs increasingly hinges on the ability to operate in critical environments without incident.
Compliance, Auditing, and Reducing the Attack Surface of Agents
For organizations deploying AI agents, whether on Frontier or other platforms, the architectural implications of this acquisition are significant. The tools from Promptfoo enable organizations to objectively assess agent behavior under adversarial constraints, document results for internal or regulatory audits, and establish measurable compliance thresholds. These requirements align closely with the emerging obligations under the European AI Act, particularly for high-risk classified systems.
Prospective customers now face the critical decision of how much reliance they are willing to place on a single platform for both deploying and auditing their AI agents. While Promptfoo’s integration into Frontier enhances the OpenAI ecosystem, it potentially diminishes the neutrality of the testing process. Organizations bound by sovereignty or independence from U.S. hyperscalers will need to evaluate whether maintaining the open-source version of Promptfoo is sufficient for ensuring autonomous evaluation capabilities.



