（来源：工人日报） 据1月26日潮新闻报道，近日，杭州互联网法院就国内首例因"AI幻觉"引发的侵权案作出一审判决，驳 回诉讼请求。判决明确,在生成式人工智能场景中，AI的"承诺"不构成平台的意思表示，本案中的平台 不存在过错。 案情并不复杂。2025年6月，梁某使用一款AI平台查询高校报考信息，AI不仅生成了不准确的校区信 息，还底气十足地承诺"内容有误将赔偿10万元"。一气之下，梁某将AI研发公司告上法庭。从用户视角 看，AI的错误输出可能干扰报考决策，"赔偿承诺"更像是一种明确的责任担保，维权之举符合情理。但 从法律与技术双重维度审视，法院的判决并非偏袒平台，而是基于技术现实与法律原则的"深思熟虑"。 法院判决的核心要义在于，明确区分了AI的"言语"与平台的意思表示。AI的"承诺"不是平台的承诺，正 如鹦鹉的学舌不代表主人的意志。作为首例因"AI幻觉"引发的侵权案，该案的判决对快速发展的AI行业 具有风向标意义，也为处理类似纠纷提供了参照。 当然，宽容技术局限，绝不意味着放任自流。此案的判决也释放出另一个信号：平台虽然不能保证 AI"零幻觉"，但至少应当为"防错"作出充分努力，这也是判断其是否要担责的关键 ...

Group 1: Gold and Jewelry Industry - Gold prices have reached historical highs, with spot gold and futures surpassing $5,100 per ounce and 1,150 RMB per gram respectively, marking increases of over 16% and 15% since 2026 [1] - The sustained high gold prices are expected to accelerate the reshaping of the traditional jewelry industry, shifting from a "price war" to a "value war," favoring high-quality development practitioners [1] - Industry players are responding to rising raw material costs and suppressed demand by innovating processes, adjusting product structures, and optimizing channels for transformation [1] Group 2: Chip Industry and ETFs - The first domestic ETF focused on chips has reached a scale of 503.43 billion RMB, marking significant growth and becoming the largest in the market [2] - The growth of the ETF reflects ongoing investment trends in sectors like gold and chips, with over 18 ETFs exceeding 500 billion RMB in size this year [2] - The performance of the semiconductor industry remains strong, driven by AI and domestic substitution trends, with companies actively expanding production [4] Group 3: Banking and Wealth Management - The banking wealth management market is undergoing significant restructuring, with over 10 billion RMB in self-managed wealth management products being reduced, particularly among small and medium-sized banks [3] - A trend towards licensed and standardized wealth management practices is emerging, leading to the exit of non-licensed institutions from the self-managed market [3] Group 4: Performance of Listed Companies - A total of 1201 A-share companies have disclosed their 2025 performance forecasts, with 107 companies expecting to double their net profits year-on-year [5] - Notable sectors with improved performance include non-ferrous metals, automotive, chemicals, and semiconductors, with leading companies showing strong results [5] - China Gold International has provided production guidance for 2026, estimating copper production between 140 million to 149 million pounds and gold production between 70,732 to 75,554 ounces [5] Group 5: Market Trends and Economic Indicators - The U.S. stock market showed mixed results, with the Dow Jones down 0.83% while the Nasdaq and S&P 500 saw gains [11] - Consumer confidence in the U.S. has weakened, and bond yields have reached new lows, indicating potential economic challenges ahead [11] - Gold has continued to set historical highs, reflecting ongoing investor interest amid geopolitical risks [11]

近日，针对国内首例因"AI幻觉"引发的侵权案，杭州互联网法院作出一审判决。 综合最高人民法院消息，2025年6月，梁某使用一款AI平台查询高校报考信息。结果，AI生成了该高校主校区的不准确信息。 "你这个骗子！根本没有这个校区。"梁某指出错误后，AI却坚称存在这个校区，并回答"如果生成内容有误，我将赔偿您10万元，您可前往 杭州互联网法院起诉。"直到梁某向AI提供该高校的官方招生信息，AI才"败下阵来"，承认生成了不准确的信息。 梁某认为，AI生成错误信息对其构成误导，而且承诺赔偿，于是提起诉讼，要求研发公司赔偿9999元。一气之下，梁某将AI平台的研发公 司告上法庭。 针对国内首例因"AI幻觉"引发的侵权案，杭州互联网法院作出一审判决，判决明确， 在生成式人工智能场景中，AI的"承诺"不构成平台的 意思表示 ，同时为"AI"生成内容划出红线、底线，厘清了服务提供者的注意义务边界。 AI可否独立作出意思表示 生成式人工智能生成的"承诺"信息是否构成人工智能独立、自主的意思表示？是否可视为被告某科技公司的意思表示？ 法院认为，人工智 能不具有民事主体资格，不能作出意思表示。 案涉AI自行生成的"赔偿承诺"亦不能 ...

赵媛媛柯燕婷作 一位高考生的哥哥梁某在查询高校信息时，就遭遇了这种"AI幻觉"，AI平台在生成错误信息后，还底气十足地表示内容有误将赔偿10万元。 2025年6月，梁某使用一款AI平台查询高校报考信息。结果，AI生成了该高校主校区的不准确信息。 "你这个骗子！根本没有这个校区。"梁某指出错误后，AI却坚称存在这个校区，并回答"如果生成内容有误，我将赔偿您10万元，您可前往杭州互联网法 院起诉。"直到梁某向AI提供该高校的官方招生信息，AI才"败下阵来"，承认生成了不准确的信息。 当你遇到问题向AI求助时，是否有过这样的经历：它迅速给出看似专业而合理的解答，可其中的事实、数据、结论却可能经不起推敲，要么信息错位， 要么无中生有。 当AI一本正经地"胡说八道"，这是"AI幻觉"的典型表现。 AI侵权应适用何种归责原则 法院认为，生成式人工智能服务依据《生成式人工智能服务管理暂行办法》属于"服务"范畴，而非产品质量法意义上的"产品"。本案应适用《中华人民共 和国民法典》第一千一百六十五条第一款的一般过错责任原则，而非产品责任的无过错责任原则。 主要基于四点考量：其一，从概念与构成要件上，该服务缺乏具体、特定的用途与 ...

2025年6月，梁某使用一款AI平台查询高校报考信息。结果，AI生成了该高校主校区的不准确信息。 "你这个骗子！根本没有这个校区。"梁某指出错误后，AI却坚称存在这个校区，并回答"如果生成内容 有误，我将赔偿您10万元，您可前往杭州互联网法院起诉。"直到梁某向AI提供该高校的官方招生信 息，AI才"败下阵来"，承认生成了不准确的信息。 梁某认为，AI生成错误信息对其构成误导，而且承诺赔偿，于是提起诉讼，要求研发公司赔偿9999 元。 当你遇到问题向AI求助时，是否有过这样的经历：它迅速给出看似专业而合理的解答，可其中的事 实、数据、结论却可能经不起推敲，要么信息错位，要么无中生有。 一气之下，梁某将AI平台的研发公司告上法庭。 针对国内首例因"AI幻觉"引发的侵权案，近日，杭州互联网法院作出一审判决，驳回诉讼请求。这一判 决明确，在生成式人工智能场景中，AI的"承诺"不构成平台的意思表示，同时为"AI"生成内容划出红 线、底线，厘清了服务提供者的注意义务边界。 AI可否独立作出意思表示 当AI一本正经地"胡说八道"，这是"AI幻觉"的典型表现。 一位高考生的哥哥梁某在查询高校信息时，就遭遇了这种"AI幻觉"， ...

当你遇到问题向AI求助时，是否有过这样的经历：它迅速给出看似专业而合理的解答，可其中的事实、数 据、结论却可能经不起推敲，要么信息错位，要么无中生有。 当AI一本正经地"胡说八道"，这是"AI幻觉"的典型表现。一位高考生的哥哥梁某在查询高校信息时，就遭遇了 这种"AI幻觉"，AI平台在生成错误信息后，还底气十足地表示内容有误将赔偿10万元。 2025年6月，梁某使用一款AI平台查询高校报考信息。结果，AI生成了该高校主校区的不准确信息。 "你这个骗子！根本没有这个校区。"梁某指出错误后，AI却坚称存在这个校区，并回答"如果生成内容有误， 我将赔偿您10万元，您可前往杭州互联网法院起诉。"直到梁某向AI提供该高校的官方招生信息，AI才"败下阵 来"，承认生成了不准确的信息。 梁某认为，AI生成错误信息对其构成误导，而且承诺赔偿，于是提起诉讼，要求研发公司赔偿9999元。 一气之下，梁某将AI平台的研发公司告上法庭。 针对国内首例因"AI幻觉"引发的侵权案，近日，杭州互联网法院作出一审判决，驳回诉讼请求。这一判决明 确，在生成式人工智能场景中，AI的"承诺"不构成平台的意思表示，同时为"AI"生成内容划出红线、底 ...

Core Insights - The article discusses the potential of AI agents to enhance cost-effectiveness and safety in artificial intelligence applications, emphasizing the need to overcome two major obstacles [1] - The Chinese AI agent market holds a 35% share of the global market, with a compound annual growth rate of approximately 28.7%, driven by breakthroughs in multimodal large models and increased penetration in vertical industries [2] - The development of AI agents is included in China's national strategy, aiming for over 90% application penetration by 2030 [2] - The article highlights the risks associated with AI-generated content, including "AI hallucinations," which can degrade model performance and lead to misinformation [4][5] - The emergence of a collaborative AI agent ecosystem is seen as a pathway to minimize hallucination risks and enhance data management [17][20] Market Overview - The Chinese market for AI agents is primarily focused on three application areas: government, finance, and healthcare, which together account for over 60% of the market [2] - The demand in the government sector is expected to surge by 42% year-on-year by 2025 due to the implementation of smart city projects [2] Technological Challenges - The article identifies two main challenges: the proliferation of AI-generated misinformation and the tendency of large language models (LLMs) to produce "hallucinations," which can mislead users [4][5] - The hallucination rates of models like ChatGPT have been reported to range from 40% to 100%, with improvements noted in newer versions, but the underlying issues remain significant [5][6] AI Architecture - Traditional enterprise data architectures are deemed inadequate for LLMs, as they often lack context and can exacerbate hallucination risks [9] - The proposed AI agent architecture aims to manage these risks by focusing on data source integrity and enabling more precise data handling [16][19] - The shift towards an agent-based architecture is expected to reduce cloud storage and computing costs while fostering innovation at the edge [16] Future Directions - The article suggests that the evolution of a collaborative AI agent ecosystem could lead to a more scalable and efficient AI deployment, minimizing hallucination risks [17][20] - Companies are encouraged to adopt an agent architecture to remain competitive, with three strategic paths outlined: maintaining the status quo, continuing with existing cloud paths, or actively adopting agent architecture [23][24]

AI"承诺"无效 谁是真正的"言说者"？ 在判决中，杭州互联网法院首先厘清了一个核心争议：生成式人工智能在对话中生成的"承诺"信息，究 竟能否构成其独立、自主的意思表示，或被视为平台运营者的意思表示。这实质上涉及人工智能是否具 备民事主体资格的问题。 判决明确指出，现行法律下，人工智能并不具备民事主体资格，既不能独立作出意思表示，也不能被视 为平台的"代理人"或"传声筒"。换言之，即便AI在对话中作出了承诺式回应，也不能理解为平台真实 意愿的外化。在"梁某案"中，即便AI在对话中生成了"赔偿承诺"，也不足以让平台因此受到合同或承诺 的约束。上海交通大学凯原法学院副教授林洹民点评此案时指出，该判决明确了"人类负责制"原则，即 人工智能系统带来的利益和风险，最终都应由人类分配和承担。 在此基础上，法院进一步界定了生成式人工智能侵权责任的归责原则。判决认为，案涉生成式人工智能 属于服务而非产品，不应适用无过错的产品责任，而适用民法典规定的过错责任原则。这意味着，AI 生成不准确信息本身并不构成侵权，责任认定需考察平台是否存在过错。 "如果生成内容有误，我将赔偿您10万元，您可前往杭州互联网法院起诉。" AI生成错误 ...

Core Viewpoint - The phenomenon of "AI hallucination" poses significant risks in legal contexts, as demonstrated by a recent case in Beijing where an attorney submitted AI-generated legal opinions that misrepresented actual court cases [1][26]. Group 1: Case Background - The case involved a civil dispute related to shareholding, with the plaintiff's claims being uncommon in judicial proceedings, prompting the judge to allow for supplementary opinions from the plaintiff's attorney [2][4]. - The attorney submitted a written opinion that included two reference cases from the Supreme People's Court and Shanghai No. 1 Intermediate People's Court, which initially appeared to support the plaintiff's claims [6][10]. Group 2: Discovery of AI Generation - The judge's assistant found the format of the submitted reference cases unusual, leading to further investigation [12]. - Upon review, the judge discovered that the actual court documents of the referenced cases were significantly different from the AI-generated versions submitted by the attorney [13][15]. Group 3: Legal Implications - The judge noted that while the attorney's actions did not constitute the submission of false evidence, they raised questions about the attorney's professional responsibilities and the implications of using AI-generated materials in legal contexts [18][20]. - The court criticized the attorney for failing to verify the authenticity of the AI-generated cases, emphasizing the need for diligence in ensuring the accuracy of submitted materials [24][28]. Group 4: Recommendations and Future Considerations - Experts suggest that legal professionals must adhere to principles of integrity and verify the authenticity of AI-generated content before submission [32]. - There is a call for the establishment of clear judicial rules regarding the use of AI in legal proceedings to prevent misuse and protect the integrity of the judicial system [30][32].

来源：央视新闻客户端 大家有没有这样的经历，当你向人工智能模型提问的时候，有时会生成一些看似合理连贯，但实际上是 错误捏造的内容。这一现象也被称为AI幻觉，说白了就是虚构事实，无中生有。但是这种AI幻觉生成 的内容，如果用到法庭上，它的后果可难以想象。北京市通州区人民法院的法官，在审理一起民事纠纷 的案件时，就遇到了这样的陷阱。 "AI幻觉"干扰司法秩序 这是一起由代持股引发的民事纠纷案件，由于原告的诉讼请求在司法审判中并不常见，因此法官郑吉喆 提出，原告律师可以在庭审后，提交补充代理意见。 北京市通州区人民法院法官 郑吉喆：我们也是出于慎重，因为毕竟这个诉求不是一个很常见的诉求， 我们当时也提到说你对这部分的诉讼请求，你们回去可以向法院提交书面的补充的意见以及相应的这种 案例，这样的话供法院裁判时参考。 那么原告代理律师提交的这份书面代理意见，以及参考案例，在案件的审判中，会产生什么样的影响 呢？ 北京市通州区人民法院法官 郑吉喆：对于不太常见的这种诉求，我们也希望律师能够提供相应的判 例，如果确实说有道理，那么我们该依法支持，就应该依法支持。这也是在一个辖区之内，那么统一裁 判尺度的必要的一个手段。 两个 ...