OpenAI buys Promptfoo for AI security
OpenAI acquired Promptfoo, embedding prompt observability and security to prevent data leakage in enterprise AI deployments.
Promptfoo's platform allows developers to simulate adversarial attacks like prompt injections and data leaks within their workflows. OpenAI plans to integrate these capabilities into its Frontier platform for building "AI coworkers". Promptfoo offers tools for scanning LLM vulnerabilities, red-teaming AI, and evaluating prompts, boasting that over 25% of Fortune 500 companies use their tools. The company's technology supports integrations with development tools and CI/CD pipelines for testing and risk detection. OpenAI's CTO of B2B Applications, Srinivas Narayanan, stated that Promptfoo's expertise will help businesses deploy secure and reliable AI applications. Promptfoo's open-source CLI and library for testing LLM applications will continue to be improved by OpenAI. Promptfoo's tools provide quantitative risk scores, potentially shortening remediation cycles from days to minutes by embedding tests inside Frontier. This acquisition underscores the increasing importance of specialized red-teaming expertise in the AI landscape.
