OpenAI has acquired Promptfoo, an AI security startup that helps companies find vulnerabilities in large language models before attackers do. TechCrunch AI reports that the deal, announced Monday, will fold Promptfoo’s red-teaming and monitoring tools directly into OpenAI Frontier, the company’s enterprise platform for AI agents.
Promptfoo was founded in 2024 by Ian Webster and Michael D’Angelo. The startup built an open source interface and library that lets organizations stress-test their LLM deployments against real-world threats. More than 25% of Fortune 500 companies already use its products, according to TechCrunch AI. Despite that traction, the company had raised just $23 million and was valued at $86 million after its July 2025 round. OpenAI didn’t disclose the acquisition price.
Why This Matters
AI agents are moving from demos to production. Companies are deploying autonomous systems that browse the web, handle customer data, and execute multi-step workflows with minimal human oversight. That’s a massive attack surface.
Bad actors don’t need to hack a traditional API anymore. They can manipulate an agent through prompt injection, trick it into leaking sensitive data, or hijack an automated workflow entirely. Every frontier lab shipping agent products knows this is the core trust problem standing between them and enterprise adoption.
This acquisition is OpenAI’s clearest signal yet that it’s treating agent security as a product-level priority, not an afterthought.
What Promptfoo Brings to OpenAI
According to OpenAI’s blog post, the integration will add three key capabilities to the Frontier platform:
- Automated red-teaming – continuous adversarial testing of deployed agents
- Agentic workflow evaluation – security analysis of multi-step agent pipelines
- Runtime monitoring – real-time risk detection and compliance tracking
OpenAI also said it expects to continue developing Promptfoo’s open source tools. That’s a smart move. The open source project is what built Promptfoo’s reputation and user base in the first place. Killing it would alienate the security research community that makes these tools valuable.
The Bigger Picture
This deal fits a pattern. As AI labs race to ship autonomous agents, they’re realizing that security infrastructure can’t be bolted on later. Anthropic has invested heavily in its own safety tooling. Google DeepMind has expanded its red-teaming operations. Now OpenAI is buying rather than building.
The $23 million raise and $86 million valuation suggest Promptfoo was still early-stage. For OpenAI, this is a relatively small bet that solves a critical gap in its enterprise story. For Promptfoo, it’s direct access to the models and infrastructure where their tools matter most.
What stands out here is the speed. Promptfoo was founded in 2024, and less than two years later it’s inside one of the largest AI companies on the planet. That’s how fast the agent security market is moving.
Enterprise buyers evaluating OpenAI’s agent platform should watch for concrete details on how Promptfoo’s tools will be packaged within Frontier. The full story is available at TechCrunch AI.