An artificial intelligence system designed for public bidding has been deployed in Anhui Province, enabling faster review of tender documents and helping regulators detect irregularities, discriminatory clauses and possible bid-rigging.
The AI bidding system was first used in Hefei, the provincial capital, during an elevator procurement evaluation. According to local officials, the system analyzed a tender and produced a detailed report in 15 minutes, a task that would normally take five human experts about three and a half hours. The human review panel later examined the AI findings and approved its conclusions.
The model was developed in 2024 by the local public resource trading center in partnership with AI company iFLYTEK. Anhui Daily reported that it was trained on 290 legal and regulatory documents, along with 82.9 terabytes of data, including previous tenders and bidding records.
Shao Hua, an official from the center, said the bidding and tendering sector involved heavy workloads and was vulnerable to corruption risks, making it a suitable area for AI use.
AI Flags Questionable Tender Clauses
Since the start of 2025, the system has reviewed about 36,000 tender documents and flagged 3,264 questionable clauses, according to local officials.
These included hidden discriminatory or exclusionary terms that can be difficult for human reviewers to identify. Among 657 AI-monitored project cases later reviewed by human experts, the AI findings were unanimously adopted in 91.17 percent of cases.
In one case involving an outdoor wall repair tender, the system detected a non-compliant qualification requirement. The model found that such restrictions are generally not allowed for standard construction projects.
Collusion Detection in Bidding
Officials said one of the system’s key functions is detecting possible bid-rigging and collusion.
Traditional monitoring often relies on signals such as IP addresses or machine codes, which can be bypassed. The AI model uses semantic analysis, logical reasoning and multimodal analysis to compare documents for unusual similarities in wording, structure, images and typographical errors.
In a sewer pipeline project, the system identified similarities between two independently submitted bidding documents, including 456 matching passages, 101 identical images and 44 shared typos.
Local officials said alerts generated by the system have helped authorities investigate 324 collusion cases so far.
National Expansion Planned
Chen Daxing of Hefei’s public resource trading center said the model went through extensive training, including learning from tens of thousands of documents, fine-tuning on thousands of projects, validation through hundreds of cases and live trials.
Officials said the model is designed to reduce the risk of AI “hallucinations” by anchoring its findings to specific laws and regulations.
During a press conference in May 2025, China’s National Development and Reform Commission said Anhui’s AI-assisted bidding evaluation had reached the level of experienced experts.
In early 2026, eight departments, including the NDRC, issued a directive to expand AI use in tendering nationwide. The plan aims to cover key areas such as document review, assisted evaluation and collusion detection in selected provinces and cities by the end of 2026, and across China by the end of 2027.
Shao said AI could help improve efficiency, reduce risks and support fairness in the bidding process.
Today's E-Paper