AIUC has raised a $40 million Series A to expand a business built around testing AI agents and insuring the risks that remain. The company argues that deployment is increasingly limited not by model capability, but by uncertainty over responsibility when an autonomous system causes damage.
Its AIUC-1 standard covers agent security, safety and reliability. Testing includes attempts to trigger jailbreaks, hallucinations and data leaks, with the goal of giving companies evidence about how a system behaves outside its normal success path. AIUC says it works with companies including Cursor, Harvey, Lovable and ElevenLabs.
The model combines technical evaluation with insurance. That could give enterprise buyers a clearer path from identifying a risk to assigning financial responsibility for it. The company also expects standards to require frequent updates because agent capabilities and attack methods change much faster than traditional certification regimes.
The announcement came through an interview with AIUC’s chief executive, so its claims about the standard’s effectiveness have not been independently tested in the source. The funding nevertheless reflects growing demand for outside evaluation as agents gain access to code, business records and physical systems where failures can have consequences beyond a bad chat response.