OpenAI to acquire Promptfoo

OpenAI just added a security shop to its growing shopping cart. The company announced it's acquiring Promptfoo, a platform built specifically to help enterprises catch vulnerabilities in AI systems while those systems are still being built, not after they've already leaked something embarrassing or dangerous into production.

Promptfoo made its name doing the unglamorous work of red-teaming: poking at prompts, testing for jailbreaks, checking whether a model will spill secrets or do something it shouldn't under adversarial pressure. That's the kind of tooling that sounds boring until the week your chatbot gets tricked into recommending a competitor's product or spitting out someone's private data, and suddenly it's the only thing anyone in the legal department cares about.

The timing tracks. OpenAI has spent the last two years trying to convince banks, hospitals, and governments that its models are safe enough to sit inside serious infrastructure, and skepticism from security teams has been one of the biggest brakes on that sales pitch. Owning a dedicated vulnerability-testing platform gives OpenAI something concrete to point to besides its own internal safety claims, which enterprise buyers have learned to treat with polite distrust.

It also fits a pattern. As foundation models get folded into more critical systems, the market for tools that audit and stress-test those models is turning into its own industry, and the big labs would rather own a piece of that industry than watch startups build businesses on top of catching their mistakes. Buying Promptfoo isn't just a security purchase. It's OpenAI buying insurance, and buying a supplier who used to answer to no one but its own customers.