Early Anthropic hire, former METR COO have found a way to rein in rogue AI agents

You can now get AI agents independently certified for safety and reliability, similar to cybersecurity standards.

Artificial Intelligence Underwriting Company (AIUC), a startup founded by an early Anthropic hire and the former COO of METR, has raised $40 million in Series A funding. The company aims to address enterprise concerns about the safety and reliability of AI agents by providing a third-party audit and certification service. AIUC has developed a standard called AIUC-1, inspired by cybersecurity frameworks like SOC 2, and employs a suite of approximately 5,000 tests to evaluate AI agents for risks such as jailbreaks, hallucinations, and data leaks. Customers like Cursor, Lovable, Harvey, and ElevenLabs are already using AIUC's services, which involve AI-driven testing with human verification of final audits. This initiative seeks to provide enterprises with concrete assurances about AI agent behavior before deployment, mitigating adoption barriers related to unknown risks.

AIUC's approach, modeled after established cybersecurity auditing practices, signifies a maturing market for AI governance where verifiable safety is becoming a prerequisite for enterprise adoption. By creating a standardized testing framework and certification, the company is establishing a crucial intermediary layer between AI developers and corporate users, addressing the growing demand for predictable AI agent behavior and mitigating deployment hesitancy.

The substantial funding and early customer traction suggest a strong market appetite for AI risk management solutions. This development could encourage other players to develop similar validation services, potentially leading to a new industry standard for AI agent trustworthiness and influencing how AI models are developed and presented to businesses.