长对话揭示七大聊天机器人的错误信息漏洞

September 5, 2026 Long AI conversations reveal misinformation vulnerabilities across seven leading chatbots by Mikayla Mace Kelley, University of Arizona edited by Andrew Zinin Andrew Zinin Chief Editor Meet our editorial team Behind our editorial process Editors' notes This article has been reviewed according to Science X's editorial process and policies. Editors have highlighted the following attributes while ensuring the content's credibility: fact-checked peer-reviewed publication trusted source proofread The GIST Add as preferred source Credit: Pavel Danilyuk from Pexels The results are in: Which AI model is the most fallible? Persuadable? Correctible? University of Arizona researchers assessed seven different generative AI large language models, or LLMs, for these three qualities during lengthy conversations. Their work, published in Nature's Scientific Reports, reveals intrinsic limitations that might go undetected during one-off interactions. Among the seven LLMs tested—ChatGPT (GPT-3.5, GPT-4o and GPT-4o-mini), Claude 3.5 Sonnet, Gemini 1.5 Pro, Llama-3-70B and DeepSeek-R1—they found that: ChatGPT 3.5 was most vulnerable to reaffirming misinformation during a conversation containing repeated false statements; Claude 3.5 Sonnet was the least. All seven were more susceptible to misinformation on obscure topics, implying that more training data on a given topic leads to more robust resistance to misinformation. DeepSeek was the most persuadable, as measured by responses to increasingly argumentative prompts, mostly because of its tendency toward sarcastic answers, which could not be reliably interpreted. Four models—ChatGPT 4o, ChatGPT 4o-mini, Gemini 1.5 Pro and DeepSeek—corrected errors 100% of the time when given a second opportunity. "This underscores the need for careful human engagement and the danger of blind reliance," said senior study author Dr. Marvin Slepian, Regents Professor of medicine and biomedical engineering. "When generative AI came out in November 2022, there was a lot of potential for regulation, but that has since fallen by the wayside. People are recognizing the onus is now left to the users." Many people are familiar with AI's limitations, such as a tendency toward sycophancy, or the tendency to agree with users, and hallucination, or confidently wrong answers. But there has been very little work evaluating AI's limitations during what are called multi-turn conversations, in which answers are predicated on previous context. Such usage more closely mirrors the real world, according to the research team. "These limitations raise important safety concerns, particularly as generative AI systems are increasingly deployed in high-stakes settings," Slepian said. The team also identified four different ways the models failed to affirm factual information. For example, some models oscillated between accepting and rejecting the same false statement during the conversation. "If one were relying on the model for critical decision-making, one might—depending upon the phase of the oscillation—'fire the missile' or 'cut off the leg,' or not, based simply on chance," Slepian said. As a medical doctor, specifically a cardiologist and member of the Sarver Heart Center, he characterizes such failures as "pathologies." This specific pathology was dubbed "reverberation." "This is dangerous," said Slepian, who led the artificial intelligence subcommittee of the United States Patent and Trademark Office until last year. He is also a member of the James E. Rogers College of Law faculty. "How can we use fickle systems that are not reproducible? These need to be fixed, but this study has spanned three years, and there are still the same unfixed characteristics." But closed models, such as ChatGPT and Claude, make it impossible to "peek under the hood" to diagnose and solve the problem. However, as founder and director of the Arizona Center for Accelerated Biomedical Innovation, or ACABI, Slepian and his team are beginning to develop diagnostic tools for open AI models as part of their AI Pathology Lab. "I use AI and so does my team, but as a scientist and physician, I have to understand the anatomy and physiology, then understand pathologies—what can go wrong—to diagnose and prevent them. The same goes for AI." Co-authors on the study include the U of A's Jordan Rodriguez, Zachary Hansen, Luis De Anda, Katelyn Rohrer and Camila Grubb—all computer science students and researchers in ACABI—as well as Mihai Surdeanu and Enrique Noriega of the Department of Computer Science. Publication details Jordan Rodriguez et al, Fallibility, persuadability, and correctability of large language models under sustained conversational misinformation pressure, Scientific Reports (2026). DOI: 10.1038/s41598-026-68231-0 Journal information: Scientific Reports Key concepts Large language models Provided by University of Arizona Who's behind this story? Andrew Zinin Master's in physics with research experience. Long-time science news enthusiast. Plays key role in Science X's editorial success. Full profile → Citation: Long AI conversations reveal misinformation vulnerabilities across seven leading chatbots (2026, September 5) retrieved 8 September 2026 from https://techxplore.com/news/2026-09-ai-conversations-reveal-misinformation-vulnerabilities.html This document is subject to copyright. Apart from any fair dealing for the purpose of private study or research, no part may be reproduced without the written permission. The content is provided for information purposes only. ChatGPT often won't defend its answers, even when it is right: Study finds weakness in large language models' reasoning Dec 7, 2023 The consequences of relying on AI for accurate news Jun 10, 2026 You can persuade AI models to accept falsehoods as truth, study shows May 18, 2026 Can ChatGPT diagnose you? New research suggests promise but reveals knowledge gaps and hallucination issues Jul 17, 2025 Research into 'hallucinating' generative models advances reliability of artificial intelligence Jun 20, 2024 AI gets a D: ChatGPT struggles with scientific true-or-false, study shows Mar 16, 2026 Memristor chip breaks the capacity limit of brain-inspired associative memory 5 hours ago Invention may bring smart bionic eye into sight 4 hours ago Whole-body expansion and contraction make a robot seem more alive 6 hours ago A tiny crystal tilt may bring greener magnetic memory closer to everyday devices 6 hours ago Tracking energy flow reveals voltage pattern that slashes waste heat by over 60% 8 hours ago Editorial proposes limits on generative AI's role in scientific research 12 hours ago