OpenAI is working to make AI systems safer for companies that rely on them in daily operations. The company has agreed to acquire Promptfoo, a startup that builds tools used to test and secure large language models.

The move comes at a time when businesses are increasingly turning to AI agents to assist with routine tasks and internal workflows.

Promptfoo’s technology will become part of OpenAI Frontier, the company’s enterprise platform designed to help organisations build and manage AI assistants. These assistants are often used to support teams and automate different parts of business processes.

As AI tools begin to take on more real-world responsibilities, companies are paying closer attention to how reliable and secure these systems are. The acquisition suggests OpenAI wants to strengthen those safeguards before AI agents become even more common in the workplace.

Promptfoo was founded in 2024 by Ian Webster and Michael D’Angelo. The startup focuses on helping developers identify weaknesses in AI models during the development stage, before the systems are widely deployed.

Its platform allows teams to run tests that mimic potential attacks or tricky prompts. This helps companies understand how an AI model might react if someone tries to manipulate it or push it beyond its intended limits.

According to Promptfoo, its testing tools are already being used by more than a quarter of Fortune 500 companies, reflecting the growing demand for ways to evaluate AI systems more carefully.

The financial terms of the deal have not been disclosed. However, data from PitchBook shows that Promptfoo has raised around $23 million since it was founded and was valued at roughly $86 million following a funding round in July 2025.

After the technology is integrated into OpenAI Frontier, companies using the platform will be able to run automated security tests on their AI agents. These checks can review how agents behave within workflows and flag potential risks before the systems are put into production.

OpenAI said the integration will also support automated “red-teaming”, a testing approach where systems are intentionally placed in difficult or adversarial scenarios to uncover hidden vulnerabilities.

In a blog post announcing the acquisition, the company noted that as AI assistants become more involved in business operations, organisations will need clearer ways to evaluate their behaviour, maintain compliance, and track activity over time.

With Promptfoo’s tools built into Frontier, businesses will be able to run security tests directly inside the platform. The system can identify issues such as prompt injection attempts, efforts to bypass safety rules, possible data leaks, or the misuse of connected tools.

The idea is to give organisations better insight into how their AI systems behave and to catch potential problems early, before those systems are deployed in real-world environments.

Also Read: OpenAI Introduces Age Prediction on ChatGPT to Strengthen Teen Safeguards