OpenAI announced on March 9, 2026, that it had agreed to acquire Promptfoo, an open-source platform for evaluating and security-testing AI applications. OpenAI said it plans to integrate Promptfoo’s technology into its Frontier enterprise platform after the deal is finalized; the announcement said closing remained subject to customary conditions, so the planned capabilities should not be read as already launched.
What does Promptfoo do?
Promptfoo is software, not a physical product: a command-line interface (CLI) and library developers can use to evaluate and red-team applications built with large language models (LLMs). Its tools cover prompts, models, agents, and retrieval-augmented generation (RAG) systems, with security and vulnerability scanning, code scanning, and support for CI/CD workflows. The project also supports multiple model providers.
In practical terms, teams can use evaluations to check whether an AI application behaves as intended, and security tests to probe how it responds to risky or adversarial inputs. OpenAI’s red-teaming guidance distinguishes the two: evaluations measure intended behavior, while red teaming looks for failure modes triggered by adversarial, abusive, or unexpected inputs. They complement each other rather than replace one another.
Why is OpenAI acquiring Promptfoo?
OpenAI says the planned combination is intended to help enterprises find and address risks in AI applications and agents during development. Its announcement names three intended areas for Frontier:
The Tool Desk
Outbyte PC Repair FREERepair Windows errors before they cause bigger problemsFix Now →Outbyte Driver Updater FREEScan for outdated or missing drivers - takes under a minuteDriver Scan →#1 Best Overall
- Automated security and safety testing: testing applications for risks such as prompt injection, jailbreaks, data leaks, misuse of tools, and agent behavior outside policy.
- Development-workflow integration: bringing security testing and evaluation into the processes teams use to build and ship AI systems.
- Reporting and traceability: giving organizations records and oversight mechanisms to support accountability.
These are OpenAI’s stated plans, not confirmation that the integrations were available when it announced the agreement. The announcement did not disclose a purchase price or other financial terms.
Will Promptfoo remain open source?
Promptfoo’s co-founders said the project would remain open source and that the company would continue serving users and customers. Its current repository identifies Promptfoo as part of OpenAI and says it remains open source under the MIT license. That repository wording indicates the project’s organizational affiliation and license, but does not state when the acquisition legally closed.
Promptfoo reported in 2026 that more than 350,000 developers had used the product, including 130,000 monthly active developers, and that teams at more than 25% of Fortune 500 companies relied on it. These are company-reported adoption figures, not independently audited results. OpenAI separately described adoption among over 25% of Fortune 500 companies.
How does AI red teaming differ from evaluations?
| Approach | What it tests | Why teams use it |
|---|---|---|
| Evaluation | Whether a system meets expected behavior or performance criteria. | To measure how well the application works against defined expectations. |
| Red teaming | How a system responds to adversarial, abusive, or unexpected inputs and other potential failure modes. | To uncover risks that ordinary evaluation may not reveal. |
The approaches work best as complementary checks: evaluations measure expected behavior, while red teaming deliberately probes for weaknesses. OpenAI advises that its Red Teaming service be used only to test code or assets the customer owns or is expressly authorized to test.
Do these 3 things before closing this tab:
1Clear out junk files and repair common Windows errors2Fix the driver behind crashes, sound loss and screen glitches3Repair Windows errors before they cause bigger problemsRank #3
What should enterprise teams take from the announcement?
The acquisition agreement points to a closer connection between AI security testing and the development workflows used to deploy agents. For teams assessing their current process, the announcement’s named risks provide a practical checklist:
- Can testing expose attempts to override instructions through prompt injection or jailbreaks?
- Can it identify unintended data exposure and inappropriate use of tools?
- Can teams detect agent actions that violate their policies?
- Can results be tracked and reviewed by the people responsible for oversight?
Promptfoo’s developer-oriented software and OpenAI’s enterprise red-teaming service are distinct ways to approach this work. The official documentation describes OpenAI Red Teaming as available to enterprise customers; it does not establish that it is the same product as Promptfoo or that it replaces local and CI/CD testing workflows.
Quick Recap
Best Value
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
