Patronus AI has secured $50 million in a Series B funding round as the startup looks to accelerate development of simulated digital environments designed to test and improve the reliability of autonomous AI agents.
The San Francisco-based company announced the funding on Friday, bringing its total capital raised to $70 million. The round was led by Greenfield Partners, with participation from Notable Capital, Lightspeed Venture Partners, Datadog, Samsung Ventures and several prominent AI industry figures.
Founded in 2023 by former Meta AI researchers Anand Kannappan and Rebecca Qian, Patronus AI is developing what it calls Digital World Models. These are highly realistic simulated environments that replicate websites, enterprise software and complex digital workflows, allowing AI agents to practice tasks and learn from mistakes before being deployed in real-world settings.
As AI systems evolve beyond conversational chatbots into agents capable of software engineering, financial analysis and other long-horizon workflows, ensuring they behave reliably has become one of the industry’s biggest technical challenges.
Traditional AI benchmarks often measure performance using static datasets, but they provide limited insight into how autonomous agents respond to unpredictable situations. Patronus AI believes realistic simulations offer a better way to evaluate and train these systems, similar to how autonomous vehicle developers use virtual environments to prepare self-driving cars for rare and dangerous scenarios.
Within these digital worlds, AI agents are trained using reinforcement learning, receiving rewards for successful decisions while being penalized for failures. The approach also helps expose shortcuts and unintended behaviours that conventional testing may overlook.
The company said demand for its platform has grown rapidly, with revenue increasing fifteenfold over the past year. Patronus AI says its technology is already being used by nearly every major frontier AI laboratory alongside numerous startups, with early deployments focused on software engineering and financial applications.
Alongside the funding announcement, the company unveiled a preview of its first Digital World Model. Built using language diffusion technology, the model is designed to predict realistic digital environments and support AI agents performing complex workflows that may last for hours, days or even weeks inside simulation.
Patronus AI believes this shift from static training datasets to dynamic simulation environments represents the next stage of AI development, allowing models to build experience through repeated practice in safe, controlled settings.
“The next phase of LLM training will be defined by simulations, carefully designed environments where agents can practice, fail and learn from long-horizon tasks,” the company said.
The new funding will be used to expand Patronus AI’s research team, grow its go-to-market operations and invest in the computing infrastructure needed to scale its Digital World Models.
As AI agents move closer to deployment across industries including healthcare, finance and enterprise automation, robust pre-deployment testing is increasingly viewed as a critical safeguard. Patronus AI is positioning its simulation platform as essential infrastructure for building more reliable and trustworthy AI systems.