Patronus AI Secures $50M Series B for Simulated Environments That Test AI Agents at Scale

Patronus AI, a startup that builds simulated digital environments for evaluating the safety and reliability of AI agents, has raised $50 million in a Series B round led by Greenfield. The company, founded by former Meta AI researchers, has now raised $70 million in total funding.

The round arrives as AI agents evolve from simple question-answering bots to autonomous systems that execute multi-step tasks, such as booking travel or managing supply chains. Patronus AI's platform lets companies run those agents through thousands of adversarial scenarios inside a simulated world before releasing them into production.

How the simulators work

Rather than testing an agent against a static dataset, Patronus AI creates interactive, digital environments that mimic real-world complexity. The agent interacts with fake APIs, customer accounts, and data stores, and the platform measures whether it follows instructions, avoids data leaks, and recovers from errors.

Key features of the platform include:

• Custom scenario generation that adapts to the specific agent being tested, covering edge cases the developer may not have anticipated.
• Automated scoring of agent outputs against safety and accuracy benchmarks, with detailed failure logs.
• Support for agents built on OpenAI, Anthropic, Google, and open-source models, with no vendor lock-in.
• Integration into CI/CD pipelines so testing runs automatically with each deployment.
• A growing library of pre-built risk scenarios for regulated industries such as healthcare, finance, and insurance.

Enterprise demand and what comes next

Patronus AI reported that revenue more than tripled over the past year as enterprises from Fortune 500 companies to early-stage startups looked for a systematic way to verify agent behavior. Investors describe demand as intense, with companies worried about the reputational and regulatory risk of a rogue agent.

The company plans to use the new capital to expand its engineering team, add support for agent teams that work collaboratively, and build out its library of domain-specific evaluation simulators. It will also open an office in New York to be closer to financial services clients.

No timeline was given for profitability. Patronus AI expects the agent testing market to grow quickly as regulators in the EU and California move toward requiring third-party audits of high-risk AI systems. The company position its simulators as a way for enterprises to get ahead of those mandates while keeping agents useful.