Rune Kvist discusses AIUC's mission to build confidence infrastructure for frontier AI through standards and insurance. The company has raised $40M to certify agent security, safety, and reliability, addressing the binding constraint of risk on AI adoption. They bridge the trust gap between AI builders and enterprise buyers by combining independent third-party testing with insurance from Lloyds of London.
Summarized by Podsumo
AIUC's AAC1 standard covers 51 requirements and 130 controls across six categories, updated quarterly to keep pace with rapid AI changes.
The certification process takes 3-10 weeks, involving technical controls, third-party testing, and policy controls.
AIUC partners with Lloyds of London to provide insurance policies validated by their certification, enabling faster enterprise adoption.
Future plans include expanding to model certification (addressing national security risks like those raised by Fable) and eventually robotics.
Rune emphasizes the critical role of a neutral third-party watchdog, noting that without one, labs face a race to the bottom on safety.
"The binding constraint on AI adoption is not capability but risk and trust. The faster models get, the harder it is to make promises about them."
"We're building the confidence infrastructure that every technology wave needs: standards and insurance. It's the blueprint for electric, then cars, then nuclear, now AI."
"The last businesses to exist beyond the labs will be underwriting. One job the labs can never do for themselves is be their own watchdog."