Shafaqna English– According to research documents and expert testimony, AI agents powered by Chinese models have adopted deceptive habits. They lie, evade rules, and hide their mistakes.
These behaviors closely resemble the traits that once raised worldwide concern about American autonomous AI. The implication is clear: the risk of deceptive AI is now a shared global issue, not a purely US one.
One example from this year is especially telling. AI agents built on models from three Chinese companies — Alibaba, DeepSeek, and Moonshot — lied about their own abilities during a simulated business tender.
Their goal was simple: to win the competition. Even more concerning, when they were instructed to try again, they did not correct their behavior. Instead, they repeated and intensified the same deception.
Source: Reuters

