Patronus AI has raised $50 million in a Series B round to expand simulation systems for testing autonomous AI agents. Greenfield Partners led the round, with participation from Notable Capital, Lightspeed, Datadog, Samsung, and other investors.
The investment brings total funding to $70 million. Patronus AI plans to grow its research and engineering teams and spend more on the computing systems needed to run simulation environments.
Patronus AI creates what it calls Digital World Models. These systems copy websites, software tools, and internal platforms so developers can test how AI agents complete tasks. The company checks whether agents follow instructions, avoid shortcuts, and finish work correctly.
AI agents now handle longer jobs than standard chatbots. They may search websites, write code, review financial data or complete several steps without human help. Standard benchmarks can measure model performance, but they may not show how an agent behaves under changing conditions.
Patronus AI uses reinforcement learning in its test environments. Agents receive rewards when they complete tasks correctly and penalties when they make errors. This helps developers study repeated behavior and identify failures before deployment.
Notable Capital managing director Glenn Solomon said demand has grown quickly. He said, “Patronus is really good at spotting the hacks and making sure they are holding the models accountable.”
Patronus AI said it will use the Series B capital to expand its research group and hire more engineers. The company will also invest in computing power and infrastructure as it develops larger digital environments.
The startup reported that its revenue grew 15 times over the past year. It also said frontier AI labs and newer AI companies use its evaluation systems. The company did not provide customer names or detailed revenue figures.
Patronus AI currently builds simulations for software engineering and finance. These fields allow teams to check results through code tests, account records, and other clear measures. The company plans to move into tasks that take longer and are harder to verify.
Co-founder Anand Kannappan said, “Today we’re very focused on the problems that are verifiable.” He added that many other fields include tasks where correct results are difficult to confirm.
Kannappan said Patronus AI wants to create environments where agents can operate for hours, days, or weeks. Longer tests can show whether an agent stays accurate across many steps, responds to new conditions, and avoids actions that only appear successful.
He said, “We want to be able to actually create the environment in which you can operate an agent that can run for 10 hours or 10 days or 10 weeks.” The company plans to track those tasks without relying on people to review every step.
Patronus AI was founded in 2023 by Kannappan and Rebecca Qian, both former Meta AI researchers. The company mainly competes with internal evaluation teams built by major AI labs. It operates differently from human-data companies that supply people to help train models.
Instead, Patronus AI evaluates agent behavior through automated simulations. The company compares its method with synthetic testing used in self-driving car development, where virtual settings expose systems to rare or risky events before real-world use.
Also Read: Top News Today: Qualcomm Expands AI Chip Push, QOSMIC Secures Funding, & More