We’re buying Promptfoo, an AI safety platform that helps enterprises establish and remediate vulnerabilities in AI methods throughout improvement. As soon as the acquisition is finalized we’ll combine Promptfoo’s expertise instantly into OpenAI Frontier, our platform for constructing and working AI coworkers.
As enterprises deploy AI coworkers into actual workflows, analysis, safety, and compliance turn out to be foundational necessities. Enterprises want systematic methods to check agent conduct, detect dangers earlier than deployment, and keep clear data to assist oversight, governance, and accountability over time.
The Promptfoo staff, led by Ian Webster and Michael D’Angelo, has constructed a strong suite of instruments trusted by over 25 % of Fortune 500 corporations, together with a broadly used open-source(opens in a brand new window) CLI and library for evaluating and red-teaming LLM purposes. Collectively, we’ll proceed constructing the open-source mission whereas additionally advancing the built-in enterprise capabilities inside Frontier.
“Promptfoo brings deep engineering experience in evaluating, securing, and testing AI methods at enterprise scale. Their work helps companies deploy safe and dependable AI purposes, and we’re excited to carry these capabilities instantly into Frontier.”
— Srinivas Narayanan, CTO of B2B Purposes, OpenAI
We’ll construct on a number of core capabilities for enterprises constructing brokers on Frontier:
- Safety and security testing constructed into the platform: Automated safety testing and red-teaming capabilities will turn out to be a local a part of the Frontier platform, serving to enterprises establish and remediate dangers like immediate injections, jailbreaks, knowledge leaks, device misuse, and out-of-policy agent behaviors.
- Safety and analysis built-in in improvement workflows: Frontier will deeply combine with the workflows wanted to establish, examine, and remediate agent dangers earlier, making safety a core a part of how enterprise AI methods are developed and operated.
- Oversight and accountability: Built-in reporting and traceability will assist organizations doc testing, monitor modifications over time, and meet rising governance, danger, and compliance expectations for AI.
“We began Promptfoo as a result of builders wanted a sensible strategy to safe AI methods. As AI brokers turn out to be extra linked to actual knowledge and methods, securing and validating them is more difficult and essential than ever. Becoming a member of OpenAI lets us speed up this work, bringing stronger safety, security, and governance capabilities to the groups constructing real-world AI methods.”
— Ian Webster, Co-founder and CEO, Promptfoo
We’re excited to welcome the Promptfoo staff and proceed constructing the instruments enterprises have to deploy safe, dependable AI.
The closing of the acquisition is topic to customary closing situations.