AI Underwriting Startup Raises $40M to Certify Autonomous Agents
Founded by former Anthropic and METR executives, AIUC is building a SOC 2-style safety standard for enterprise AI.
Artificial Intelligence Underwriting Company (AIUC) has raised $40 million in Series A funding to build a third-party certification layer for AI agents. The startup aims to prevent autonomous systems from "going rogue" as enterprises integrate AI into critical operational environments.
The funding round was led by Ribbit Capital, with participation from First Harmonic. This brings AIUC's total funding to $55 million, following a $15 million seed round backed by Nat Friedman (NFDG), Emergence, Terrain, and Ben Mann. Founded by early Anthropic employee Rune Kvist and former METR COO Rajiv Dattani, the company has already secured a customer base that includes ElevenLabs, Harvey, Lovable, and Cursor.
A Standard for AI Safety
To validate agent behavior, AIUC developed the AIUC-1 safety standard in collaboration with a consortium of approximately 250 security and risk leaders. The company employs a testing suite of roughly 5,000 individual tests designed to detect hallucinations, data leaks, and jailbreaks. This process culminates in a comprehensive audit report of about 100 pages, providing a technical guarantee of an agent's safety profile.
The approach mirrors the SOC 2 model used in cybersecurity, shifting the industry focus from a model's raw intelligence to its reliability. This shift aligns with recent calls from Anthropic CEO Dario Amodei for paced frontier development and the integration of third-party evaluators to manage the risks of increasingly capable models.
The Trust Gap in Enterprise AI
As companies transition from simple chatbots to autonomous agents with the power to execute actions across systems, the potential for unpredictable behavior grows. According to Rune Kvist, the paradox of AI development is that as systems become smarter, they often become harder to control and adopt.
Kvist notes that highly regulated sectors—including militaries, governments, hospitals, and banks—are not avoiding AI because of a lack of capability. Instead, they are hesitant because they cannot guarantee that a system will adhere to specific safety commitments made to their customers. AIUC is positioning itself as the "underwriting" layer that provides the necessary trust for these sectors to deploy agents at scale.
The Path to Deployment
AIUC's success depends on whether the AIUC-1 standard can gain widespread industry adoption as the benchmark for AI safety. While the company has already attracted several high-profile AI startups as clients, the broader challenge remains the volatility of frontier models, which can exhibit new failure modes as they evolve.
Observers will be watching to see if this certification model becomes a requirement for AI deployment in regulated industries, effectively creating a new industry of AI safety auditing that parallels the existing financial and cybersecurity compliance landscapes.