A new startup aims to reduce enterprise risks from uncontrolled AI agents by applying a cybersecurity-inspired audit and certification framework, backed by $40 million in Series A funding.
- AIUC develops AIUC-1, a safety certification standard for AI agents.
- The startup raised $40 million in Series A from Ribbit Capital and First Harmonic.
- Testing involves 5,000 scenarios analyzed by AI with human audit verification.
What happened
AIUC, a new startup formed by former key figures from Anthropic and METR, announced a $40 million Series A funding round led by Ribbit Capital with participation from First Harmonic. The company’s total funding to date is $55 million, including a $15 million seed round featuring investors like Nat Friedman and Emergence.
AIUC has developed an AI safety standard called AIUC-1 along with a rigorous testing service aimed at helping enterprises assess and certify the behavior of AI agents. The service runs extensive tests simulating jailbreaks, hallucinations, and data leaks, producing detailed reports on AI agent safety and reliability.
Why it matters
As AI agents become increasingly sophisticated, their unpredictability and risks grow, making it harder for institutions such as banks, hospitals, and governments to confidently deploy these technologies. Existing safety commitments often prevent organizations from adopting AI without reliable assurances of behavior and security.
By applying a cybersecurity-style auditing model tailored for AI, AIUC addresses a critical market need for third-party validation of AI agent safety. Their approach draws on input from 250 security and risk leaders, aligning certification criteria with what buyers of AI agents consider essential for risk management and compliance.
What to watch next
AIUC’s rollout of the AIUC-1 standard and its certification services will be closely watched by enterprises deploying AI technologies, as well as investors focused on AI safety and governance. The startup’s ability to scale testing and gain widespread industry adoption will be vital to its success.
The broader AI safety ecosystem may see increased collaboration or competition as other organizations like METR push frontier lab testing and calls for regulated third-party evaluators gain traction. AIUC’s integration of AI to test and analyze AI agents, with human oversight, may set new precedents for balancing automation and accountability in AI safety evaluations.