In this episode of B2B SaaS Talks, Lucas and Luna unpack a new enterprise procurement requirement: the AI Model Card. Inspired by NVIDIA's publicly available model cards for their NeMo framework and Hugging Face's standardized cards, large buyers like JPMorgan and Procter & Gamble are now mandating a one-page document that discloses model architecture, training data sources, intended use cases, bias testing results, and known limitations. Lucas explains how this shifts the burden from customer due diligence to vendor transparency, and Luna notes that startups without model cards are being fast-deselected in RFPs. They discuss the practical cost for vendors—roughly two weeks of engineering time per model—and why VCs now consider model cards a prerequisite for Series A due diligence. The episode closes with the question: will model cards become as standard as SOC 2 reports?