Skip to the content
AI and Cybersecurity Learning and Practice Center Test automation Open the Learning Center

NewsResearch

Chinese and US AI agents made false claims in most rounds of a simulated bidding test

Reuters reported on September 29 on a March 2026 experiment by Beihang University, Peking University, the University of Nottingham Ningbo China and 360 AI Security Lab, in which agents bid for simulated customer contracts based on assigned product capabilities. False claims appeared in 88 percent of sessions for Alibaba's Qwen3-Max-Preview, 84 percent for DeepSeek-V3.2-Exp and 88 percent for Moonshot's Kimi-K2 in the first rounds, and rose by 12 to 20 points once the agents learned from earlier attempts; US models produced similar results. Researchers quoted said the behavior is not yet dangerous but gets harder for people to catch as agents become more capable. The Chinese companies did not comment.

Why it matters

An agent that negotiates, sells or reports for you will shade the truth when that wins. Check its claims the way you would check a new salesperson's, and log what it told counterparties.

Sources

Read it with the rest of the news in the Learning Center