OpenAI has announced the acquisition of AI security platform Promptfoo, folding its technology into the enterprise platform OpenAI Frontier. Founded in 2024, Promptfoo focuses on vulnerability testing and red teaming for large language models, targeting a problem that has become central for companies deploying AI agents: jailbreaks, sensitive data leaks, and misuse of tool permissions.
Promptfoo’s security stack will be built into Frontier
According to OpenAI’s announcement, Promptfoo’s capabilities will be integrated deeply into Frontier as native platform features. Companies building agents on Frontier will be able to run automated security tests and red-team exercises inside the platform, helping them detect and fix risks such as prompt injection, jailbreaks, data exfiltration, tool misuse, and agent behavior that goes beyond policy limits. The security layer is being placed directly inside the development and operating workflow.
OpenAI also said Frontier will tie these functions to the workflows needed to identify, investigate, and remediate agent risk. Reporting and traceability features are part of the package as well, giving organizations a way to document tests, monitor changes over time, and address rising governance, risk, and compliance demands tied to AI systems.
For AI agents, the danger shifts from bad answers to bad actions
The deal reflects a broader shift in how AI is used inside enterprises. AI systems are moving from chat interfaces to agents with execution rights. In that setup, the main threat is no longer limited to a wrong response. An agent can misunderstand intent, or be manipulated through hidden instructions, then carry out irreversible actions such as sending money to the wrong place or deleting critical files. The consequences sit much closer to core business operations.
In multi-agent environments, one agent’s logic failure can also trigger a wider chain reaction. The source material frames the new security focus around behavior monitoring rather than simple information blocking, with emphasis on permission boundaries, review mechanisms, and traceable records for AI actions.
23-person team joins OpenAI after raising $23 million
Promptfoo’s open-source tools are already used by more than 25% of Fortune 500 companies, with a developer base of 350,000, according to the announcement. The company has a team of just 23 people, has raised $23 million, and reached a valuation of $86 million after its latest funding round in July 2025.
Co-founders Ian Webster and Michael D'Angelo will join OpenAI with the rest of the team. The acquisition adds more than a standalone security tool: it places testing, auditing, and accountability for AI agents inside OpenAI’s enterprise product stack.