Guang Zhou Ri Bao·2026-03-12 04:13Core Insights - The article discusses the upgrade of Meituan's AI model "Xingmou," which enhances food safety governance in the food delivery industry by expanding its application from kitchen recognition to various aspects such as merchant onboarding, store authenticity verification, and kitchen environment alerts [2][3] Group 1: Store Authenticity Verification - The upgraded "Xingmou" model strengthens the ability to verify store authenticity by comparing uploaded storefront images with real-time photos, effectively identifying inconsistencies and potential fraud [3] - The AI system has conducted over 250,000 storefront consistency checks, transitioning from a one-time manual review to a dynamic, ongoing intelligent audit process [3] - The model has intercepted over 450,000 suspected AI-generated storefront images during the onboarding process, addressing a significant challenge in food safety governance [3] Group 2: Video Verification - Meituan has implemented a "one-shot" video verification feature requiring new merchants to upload complete real-time videos with location information, enhancing the verification process [4] - The upgraded model can identify inconsistencies in the video footage, such as abnormal visuals and hygiene conditions in the kitchen [4] Group 3: Kitchen Environment Alerts - The AI model provides real-time alerts for kitchen environment issues, identifying risks such as unapproved practices, waste overflow, and pest presence, thus improving food safety [5] - The kitchen alert capability now covers 910,000 merchants on the Meituan platform, with daily inspections reaching 14 million, surpassing previous manual inspection frequencies [6] - The average accuracy rate of the "Xingmou" model in recognizing kitchen conditions is 94.6%, significantly higher than the industry average of 65% [6]