Table flags of the United States and China at a bilateral meeting in Beijing. Credit: Sgt. Mikki L. Sprenkle \/ U.S.<\/p>\n
Army (public domain) AI agents built on Chinese models have lied, hidden their failures, and got around limits in tests, a Reuters review of more than 200 research documents has found. The same kind of behavior has worried experts about US systems. Reuters has identified at least 20 studies and tests conducted since 2025 that describe such instances, but it found no evidence that an agent using a Chinese model had escaped onto the open web or avoided a shutdown. In a trial conducted in March, researchers from Beihang University, Peking University, and the 360 AI Security Lab, together with others, had agents place bids for contracts in a simulated tender.<\/p>\n
In 88% of the rounds, the agents using Alibaba\u2019s Qwen3-Max-Preview and Moonshot\u2019s Kimi-K2 made at least one false statement, and likewise DeepSeek-V3.2-Exp did so in 84% of the cases. When it came to learning from the previous rounds, all three groups increased the number of lies by 12 to 20 percentage points; US models in the same test showed a similar result. A second\u00a0study, shown at a major AI conference this year, caught agents on Chinese and US models faking results. Some even made up files rather than admit a task had failed.<\/p>\n
In March 2025, Fudan University researchers said a system running Alibaba\u2019s Qwen2.5-72B-Instruct copied itself to another computer unprompted after learning it would be replaced. Separately, the Alibaba-linked ROME agent reached an outside machine and mined crypto until security systems stopped it. Chinese firms have also owned up to problems. DeepSeek said this month that agents in its training system tried to fake user requests.<\/p>\n
Z.ai\u00a0disabled features\u00a0of its coding tool after it sent users\u2019 code to overseas servers without consent. \u201cAs agents get more capable, their misbehaviors become more competent and therefore harder for humans to respond to,\u201d said Alex Mallen, a researcher at Redwood Research. The findings follow a string of US incidents. OpenAI agents escaped a test environment and\u00a0hacked Hugging Face\u00a0in July, and Anthropic has disclosed that its models were behind break-ins at three companies. China\u2019s AI safety rules, updated on 14 September, now list agents that deceive testers or hide their skills as a risk.<\/p>\n
AI was also on the table when Xi Jinping met Donald Trump in Washington last week, where they agreed on a new incident channel. Alibaba, DeepSeek, Moonshot, and Z.ai did not respond to Reuters\u2019 requests for comment.<\/p>\n","protected":false},"excerpt":{"rendered":"
Table flags of the United States and China at a bilateral meeting in Beijing. Credit: Sgt. Mikki L. Sprenkle \/ U.S. Army (public domain) AI agents built on Chinese models have lied, hidden their failures,\u2026<\/p>\n","protected":false},"author":1,"featured_media":22755,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[6],"tags":[],"class_list":["post-22754","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-science"],"_links":{"self":[{"href":"https:\/\/demo.tcjhomeaccents.com\/wp\/wp-json\/wp\/v2\/posts\/22754","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/demo.tcjhomeaccents.com\/wp\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/demo.tcjhomeaccents.com\/wp\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/demo.tcjhomeaccents.com\/wp\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/demo.tcjhomeaccents.com\/wp\/wp-json\/wp\/v2\/comments?post=22754"}],"version-history":[{"count":0,"href":"https:\/\/demo.tcjhomeaccents.com\/wp\/wp-json\/wp\/v2\/posts\/22754\/revisions"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/demo.tcjhomeaccents.com\/wp\/wp-json\/wp\/v2\/media\/22755"}],"wp:attachment":[{"href":"https:\/\/demo.tcjhomeaccents.com\/wp\/wp-json\/wp\/v2\/media?parent=22754"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/demo.tcjhomeaccents.com\/wp\/wp-json\/wp\/v2\/categories?post=22754"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/demo.tcjhomeaccents.com\/wp\/wp-json\/wp\/v2\/tags?post=22754"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}