{"type":"video","version":"1.0","html":"<iframe src=\"https://www.loom.com/embed/7fb37c1f816e42fea4394933f1ddc2a3\" frameborder=\"0\" width=\"1920\" height=\"1440\" webkitallowfullscreen mozallowfullscreen allowfullscreen></iframe>","height":1440,"width":1920,"provider_name":"Loom","provider_url":"https://www.loom.com","thumbnail_height":1440,"thumbnail_width":1920,"thumbnail_url":"https://cdn.loom.com/sessions/thumbnails/7fb37c1f816e42fea4394933f1ddc2a3-41501fa32200a5b9.gif","duration":231.213,"title":"Acipta.ai Evidence and Compliance for AI","description":"This Loom explains how Acipta.ai provides cryptographically signed, byte-identical replayable AI compliance verdicts for enterprises deploying LLMs at scale. It contrasts two CSA operating documents: securing LLM backend systems and managing AI model risk, noting the common gap where teams cannot produce defensible evidence fast enough. Asipta uses three frontier models (claims include GPT-5, Gemini, and a Cloud model) with a 0.85 confidence floor and a consensus rule requiring at least two-thirds agreement; if not, it routes to a human in the loop. Each verdict is signed at write-on, timestamped using RFC 31618, and replayable for five years, with GA expected on August 23."}