OpenAI agrees to acquire Promptfoo

OpenAI has agreed to acquire Promptfoo, a fast-growing artificial intelligence security platform, in a move aimed at strengthening the safety and reliability of enterprise AI systems as companies accelerate the deployment of autonomous agents and generative models across critical operations.

The acquisition will bring Promptfoo’s specialised testing and vulnerability-detection technology into OpenAI’s enterprise environment, known as OpenAI Frontier, a platform designed to help organisations develop and operate AI-powered assistants and agents. Financial terms of the deal have not been disclosed. The transaction reflects mounting industry concern that increasingly powerful AI systems could introduce new cyber-security risks if deployed without robust evaluation frameworks.

Promptfoo develops tools that allow companies to simulate adversarial attacks against AI systems during the development phase. These tests identify weaknesses such as prompt injection, data leakage, misuse of integrated tools and manipulation of retrieval systems. By running automated “red-team” simulations against applications before they are released, developers can uncover vulnerabilities and fix them before attackers exploit them.

OpenAI plans to integrate these capabilities directly into Frontier, enabling enterprise customers to run systematic security testing on AI agents and large language model applications within their development pipelines. The approach is intended to embed safety checks into the software development lifecycle, allowing organisations to identify potential failures before systems reach production environments.

The move highlights a shift in the artificial intelligence industry, where performance breakthroughs have been followed by growing scrutiny over safety, governance and operational risks. Companies deploying AI at scale now face a widening attack surface, particularly when models are connected to databases, software tools and external data sources.

Promptfoo emerged as one of the earliest platforms focused specifically on evaluating generative AI systems. Founded in 2024, the company developed an open-source framework that allows developers to stress-test AI models using simulated malicious prompts and automated attack scenarios. The platform quickly gained traction among developers and enterprise teams seeking practical tools to secure applications built on large language models.

Adoption expanded rapidly as enterprises experimented with AI-driven automation across customer service, data analysis and internal productivity tools. According to industry figures, Promptfoo’s open-source tools have been downloaded by more than 100,000 developers, while dozens of Fortune 500 companies have deployed its enterprise platform within production systems.

Security researchers warn that generative AI introduces risks that traditional cyber-security tools are poorly equipped to address. Attackers can manipulate AI behaviour through prompt injections, exploit vulnerabilities in retrieval-augmented generation systems, or trick models into exposing confidential information. Autonomous agents that perform actions across corporate systems may also become entry points for data breaches or operational disruptions.

Promptfoo’s testing framework attempts to address these issues by automating adversarial evaluations. The software generates thousands of context-specific attack prompts designed to probe weaknesses in models, applications and integrations. Results are fed back into development workflows, enabling engineers to implement fixes and monitor vulnerabilities over time.

The acquisition signals OpenAI’s broader strategy to build a full enterprise ecosystem around its AI models, including governance, compliance and security layers. As organisations move from experimental deployments to mission-critical uses, vendors are increasingly expected to provide integrated safeguards alongside performance improvements.

Industry analysts note that the race to commercialise AI has created pressure to deliver tools that help enterprises manage risk. Regulators across several regions are also introducing frameworks requiring companies to document how AI systems are tested and monitored, particularly when they influence business decisions or customer interactions.

Promptfoo’s leadership team, including co-founders Ian Webster and Michael D’Angelo, will continue working on the technology following the acquisition. The platform’s open-source community is also expected to remain active, with development continuing alongside the enterprise integration into OpenAI’s infrastructure.

OpenAI has expanded through several acquisitions and partnerships as it positions itself as a provider of enterprise-grade AI platforms. The company has sought to broaden its capabilities beyond language models, adding developer tools, monitoring systems and operational infrastructure intended to support large-scale deployment of AI services.

Integration of Promptfoo’s technology into Frontier is expected to enable continuous security evaluation of AI agents, allowing organisations to test behaviour, document vulnerabilities and maintain audit trails for compliance purposes. The framework also supports monitoring after deployment, helping enterprises track whether system updates introduce new weaknesses.

Cyber-security specialists increasingly view such evaluation systems as essential to the future of AI adoption. As models become more capable and autonomous, the need to identify risks before they affect real-world systems has become a central concern for developers, regulators and corporate users alike.

Growth in the AI security sector reflects the industry’s attempt to address these challenges. Start-ups focused on testing, governance and risk management have attracted significant investment as enterprises seek to deploy artificial intelligence while maintaining control over safety and reliability.