AI 'new colleagues' launched for CIIE oversight: Food Safety AI Doctor, advertising inspection, and smart elevator codes

Restaurant hygiene can be "checked with a single photo," elevator safety "known with a single scan," and whether a livestream violates regulations can be identified "in seconds"... At the 9th China International Import Expo (CIIE), the comprehensive service and support command platform will be even "smarter." Reporters today accompanied the market regulation authorities on a site visit and learned that during this year's CIIE, "new AI colleagues" will join the supervision of every link—from dining tables and elevators to booth livestreams—using digital means to build multiple lines of defense for food safety, special equipment safety, and expo order. Food safety checks automatically extract frames. At this year's CIIE, there are 55 catering units and 4 collective meal-supply enterprises in the venue. How can food safety supervision be made more efficient and precise? At the CIIE service and support comprehensive command center of the Municipal Market Supervision Administration near Hall 7, staff demonstrated the newly upgraded "Food Safety AI Doctor." The biggest highlight of "AI Doctor" is its embedding of the "Food Safety Smart Eye (check with a single photo)" agent, which can conduct intelligent assessment, quantitative scoring, compliance guidance, and full-process traceability for catering enterprises, turning kitchen images into quantifiable and evaluable data. When support staff use a mobile phone to photograph kitchen operation scenes, the system can automatically identify risks, output a problem list, and propose rectification suggestions. For the 4 CIIE collective meal-supply units distributed across 3 jurisdictions, the market regulation authorities carry out full-chain intelligent risk monitoring, evaluation, and early warning across 8 links including raw material procurement and cooking and processing, and meal preparation and supply, forming a closed-loop management mechanism of "problem discovery—follow-up rectification." During the interview, a kitchen manager at a meal-supply enterprise received an AI warning while conducting a self-inspection through the mini-program, indicating that cleaning tools were too close to the food operation area and there was a risk of cross-contamination. The merchant immediately relocated and rectified the issue, eliminating the hidden danger. It is worth noting that the system has also promoted the upgrade of "transparent kitchens." There are a total of 145 mobile video surveillance signals from catering units in the venue, all of which are now connected to the command platform. During the three processing periods of breakfast, lunch, and dinner each day, the system automatically extracts and analyzes frames from surveillance footage twice, scoring the three major elements of environmental hygiene, process standardization, and personnel hygiene, and assigning three risk levels: low, medium, and high. During the interview, reporters saw that the system automatically extracted real-time surveillance footage from a certain catering enterprise and found that vegetable baskets that should have been placed on tables were piled on the floor, with a large amount of food residue and garbage scattered on the ground. The system immediately provided AI verification results and rectification suggestions, and the merchant only passed after completing rectification and uploading photos. "For high-risk units scoring below 80, regulators immediately visit to inspect and provide guidance; for medium-risk units scoring 80 to 95, daily follow-up rechecks are conducted; for low-risk units scoring above 95, off-site regulatory guidance is the main approach," the regulator introduced. It is reported that the mini-program was launched in March this year and has been continuously fed data for iterative optimization; its overall recognition accuracy now exceeds 90%. "The system unifies regulatory evaluation standards, avoids subjective judgment differences among different regulators, and improves inspection efficiency. In the future, we will gradually expand the system's coverage and promote it to catering enterprises across society," said Qiu Congqian, deputy director of the Office of the Leading Group for Market Regulation Service and Support of the 9th CIIE. Ad inspections are more precise. Advertising inspections have always been a key focus of the CIIE's comprehensive market regulation support. In previous years, inspection personnel had to enter the exhibition halls in advance and check promotional materials such as posters, display boards, and electronic screens one by one. This year, the municipal market regulation authorities launched the "Advertising AI Intelligent Comprehensive Inspection System." Hall inspectors log into the system via mobile phone, create work orders, and upload photos; AI then automatically compares them against the advertising corpus and flags suspected violations, while the back end simultaneously generates risk assessment references. Inspectors review and manually verify, moving the advertising compliance checkpoint forward. For the growing number of booth livestreams in the venue, intelligent supervision is especially effective. At last year's CIIE, many exhibitors had already promoted products through livestreaming. In response to this trend, the system has added a livestream information collection module that can register and file information such as livestream platforms, MCN agencies, streamers, and product categories, while carrying out livestream inspections in parallel, reflecting prior guidance, in-process reminders, and standardized guidance. According to reports, the "second-level" recognition is backed by years of practical experience at exhibitions. The system integrates the advertising corpus compliance database of the China Advertising Association and establishes high, medium, and low three-level risk assessment standards tailored to CIIE scenarios, precisely focusing on high-frequency hidden risks of advertising violations. In addition, the system can simultaneously collect and analyze inspection data. The back-end command screen in real time summarizes and updates the total number of advertising points in each exhibition hall, inspection progress, number of problematic advertisements, proportion of medium- and high-risk cases, and high-incidence industry trends, allowing the command center to conduct dynamic scheduling of inspections across the entire venue. Compared with last year, full-coverage inspections that originally required a whole day or even continued until just before the venue opened can now be completed in about 7 hours, greatly improving inspection efficiency. Elevator safety is known with a single scan. The National Exhibition and Convention Center has 520 pieces of special equipment, including 424 elevators. After the expo opens, elevator passenger flow in the venue is high, making them a top priority for inspections; in previous CIIE years, full-coverage inspections were conducted more than twice a day. At this year's CIIE, elevator safety support has also achieved full-process optimization of "scan-code verification—image collection—AI analysis—problem rectification." AI-assisted frame extraction scans on-site footage to identify potential hidden dangers such as equipment operation and component status, promoting "early discovery, early warning, and early handling" of risks. Inspection personnel only need to use a mobile phone to scan the "Smart Elevator Code" in a prominent elevator location to retrieve equipment registration information, fault records, maintenance records, previous inspection records, and closed-loop rectification status, and generate a new work order. After a law enforcement recorder captures the "Smart Elevator Code," it can also automatically associate it with the work order; AI displays the analysis results in real time, and after manual confirmation, a handling record is formed, greatly improving inspection efficiency. In addition, the system has achieved full coverage of remote monitoring for 424 elevators in the venue and is connected with the systems of maintenance units. Combining escalator passenger flow monitoring, inspection video comparison, and back-end AI algorithms, precise maintenance can be implemented, driving the upgrade of supervision from "human defense" to "technical defense."