Anthropic指控阿里巴巴非法提取Claude AI模型能力Anthropic Accuses Alibaba of Illicitly Extracting Claude AI Model Capabilities
美国人工智能巨头Anthropic在一封致美国国会议员的信中，指控中国科技与电子商务巨头阿里巴巴集团对其Claude AI模型进行了“公然”且“非法”的能力提取。Anthropic在信中详细指出，与阿里巴巴相关的操作员使用了数千个虚假账户，与Claude模型进行了近2900万次交流，通过所谓的“蒸馏攻击”提取更强AI模型的答案来训练较弱的模型，这是迄今为止已知规模最大的此类能力提取活动。 在这封日期为6月10日、致美国参议员蒂姆·斯科特（Tim Scott）和伊丽莎白·沃伦（Elizabeth Warren）的信中，这家总部位于旧金山的公司敦促美国国会采取行动，惩罚发起此类攻击背后的相关公司，并进一步加强防范措施以防止美国的核心技术被窃取。受此指控消息的直接影响，阿里巴巴在纽约证券交易所交易的美国存托凭证（ADR）在周三下午交易中出现下跌，跌幅超过3%，触及每股99.10美元的盘中低点。
US artificial intelligence giant Anthropic has accused Chinese e-commerce and technology conglomerate Alibaba Group of "brazenly" and "illicitly" extracting capabilities from its Claude AI model, according to a letter sent to US lawmakers. In the letter dated June 10, which was addressed to US Senators Tim Scott and Elizabeth Warren, the San Francisco-based company stated that operators linked to Alibaba used thousands of fraudulent accounts to conduct nearly 29 million exchanges with Claude. Anthropic described this campaign as the largest known "distillation attack" of its kind, where answers were extracted from a stronger model to train a weaker one. In light of these findings, Anthropic urged the US Congress to penalize the companies behind such attacks and strengthen measures to prevent the theft of US technology. Following the disclosure of these allegations, Alibaba's American depositary receipts (ADRs) fell more than 3% on Wednesday afternoon in New York, hitting a session low of $99.10.
01Anthropic指控阿里巴巴非法提取AI模型能力Anthropic Accuses Alibaba of Illicit AI Capability Extraction
总部位于旧金山的人工智能初创公司Anthropic近日向美国国会参议员递交指控信，公开指责中国科技与电子商务巨头阿里巴巴集团对其旗下的Claude AI模型实施了大规模的“蒸馏攻击”。在这封日期为6月10日的信函中，Anthropic明确指控阿里巴巴以“公然”且“非法”的方式提取了Claude模型的技术能力。这一指控将两家科技巨头在人工智能技术安全领域的争端推向了前台。
San Francisco-based artificial intelligence startup Anthropic has formally accused Chinese technology and e-commerce giant Alibaba Group of executing a massive "distillation attack" to extract capabilities from its Claude AI model. In a letter dated June 10 addressed to US lawmakers, the company alleged that Alibaba "brazenly" and "illicitly" targeted its proprietary technology. This formal complaint highlights growing disputes between major American and Chinese technology firms over intellectual property and AI model security.
根据相关报道，Anthropic在信中将此次事件描述为该公司迄今为止所发现的“已知最大规模的能力提取活动”。该公司声称，与这家在纽约证券交易所上市的中国互联网巨头相关的操作人员，旨在利用Claude这一更强大AI模型所生成的输出结果，来训练和提升自身较弱的AI模型。目前，媒体已联系阿里巴巴请求置评。
According to the correspondence, Anthropic characterized the operation as the largest known campaign of its kind directed against its systems. The AI developer asserted that operators linked to the New York Stock Exchange-listed Alibaba sought to harvest high-quality data from Claude, a more advanced model, to train and refine Alibaba's own weaker AI systems. While the allegations have sparked significant attention, media outlets have reached out to Alibaba for comment.
Anthropic与阿里巴巴：AI竞争与地缘政治摩擦时间线Anthropic and Alibaba: AI Competition and Geopolitical Friction Timeline
展示在Anthropic指控阿里巴巴非法提取模型能力之前，双方在AI估值、模型发布、美国黑名单制裁及出口管制方面的地缘政治与技术演变背景。
Shows the geopolitical and technological evolution involving AI valuations, model releases, US blacklisting, and export controls leading up to Anthropic's accusations against Alibaba.
Anthropic在新一轮融资中筹集了650亿美元，估值达到9650亿美元，超越OpenAI成为全球最有价值的AI初创公司。
Anthropic raised $65 billion in a funding round, pushing its valuation to $965 billion and leapfrogging OpenAI as the world's most valuable AI startup.
美国国防部更新中国军事企业名单，将阿里巴巴、百度和比亚迪重新列入黑名单，禁止其获得美国国防合同。
The US Department of Defense updated its list of Chinese military companies, restoring Alibaba, Baidu, and BYD to the blacklist and barring them from US defense contracts.
Anthropic向企业客户和付费订阅用户开放了其最强大的Mythos级别AI模型Claude Fable 5。
Anthropic released its most powerful Mythos-class AI model, Claude Fable 5, to enterprise customers and paid subscribers.
为遵守美国政府旨在限制外国国民访问的紧急出口管制指令，Anthropic被迫停用了其最新发布的Fable 5和Mythos 5模型。
Following an emergency US export control directive to block foreign nationals, Anthropic disabled access to its newly released Fable 5 and Mythos 5 models.
中国商务部和财政部宣布对数十家美国企业实施双用途出口管制与政府采购禁令，以反制美国此前将阿里巴巴等列入军事黑名单的行为。
China's Ministry of Commerce and Ministry of Finance announced trade restrictions and procurement bans on US firms, retaliating against the US blacklisting of Alibaba and other Chinese tech giants.
02攻击细节：数千虚假账户与近2900万次交互Details of the Attack: Thousands of Fake Accounts and 29 Million Exchanges
在其正式指控中，Anthropic提供了关于此次行动具体规模的详细细节。这家人工智能公司声称，与阿里巴巴相关的操作人员系统性地创建并控制了数千个虚假账户以访问其系统。通过这些虚假账号，操作人员得以与Claude聊天机器人进行了近2900万次独立交流。这种方式使他们能够规避常规的使用限制，并收集到极其庞大的专有数据。
In its formal complaint, Anthropic provided specific details regarding the scale of the alleged operation. The artificial intelligence company asserted that operators associated with Alibaba systematically set up and controlled thousands of fraudulent accounts to access its systems. Through these fake profiles, the operators managed to conduct almost 29 million individual exchanges with the Claude chatbot, allowing them to bypass normal usage barriers and gather a massive volume of proprietary data.
Anthropic将这种大规模的数据收集活动归类为“蒸馏攻击”。在人工智能领域，蒸馏攻击是指外部操作者系统性地向一个更先进、更强大的AI模型发送请求并提取其输出解答。随后，这些收集到的数据会被用作训练样本，来训练和改进另一个性能较弱的AI模型，从而使发起攻击的实体能够以极低的研发成本复制目标模型的技术能力。
Anthropic classified this extensive data-gathering campaign as a "distillation attack." In the field of artificial intelligence, a distillation attack refers to a method where external actors systematically query a highly advanced and stronger AI model to extract its answers. The collected outputs are then used as training data to improve and refine a separate, weaker AI model, allowing the attacking entity to replicate the target model's capabilities at a fraction of the development cost.
03Anthropic呼吁国会采取惩罚与防范措施Anthropic Calls for Congressional Action and Penalties
Anthropic在致美国参议员蒂姆·斯科特（Tim Scott）和伊丽莎白·沃伦（Elizabeth Warren）的信中，对这种技术提取行为表达了深切关切。这家总部位于旧金山的公司明确敦促美国国会采取行动，惩罚发起此类“蒸馏攻击”背后的相关公司。Anthropic强调，国会需要制定并实施更强有力的政策，以防止美国的技术和知识产权被外部实体非法窃取。
In its letter addressed to US Senators Tim Scott and Elizabeth Warren, Anthropic expressed deep concern over the unauthorized extraction of its technology. The San Francisco-based artificial intelligence company explicitly urged the US Congress to take action to penalize the entities responsible for orchestrating these distillation attacks. Anthropic emphasized the critical need for lawmakers to implement stronger protective measures, arguing that robust intervention is necessary to prevent valuable US technology and intellectual property from being stolen by foreign competitors.
在这封日期为6月10日的信函中，Anthropic呼吁立法者加大防范力度，提升美国科技行业的整体防御能力。该公司指出，针对先进AI模型的非法能力提取活动不仅损害了开发者的利益，也对技术安全构成了挑战。通过向这两位参议员发出呼吁，Anthropic旨在推动国会出台具体措施，遏制此类利用虚假账户规避限制并系统性获取专有数据的行为。
The correspondence, dated June 10, outlines Anthropic's appeal for a coordinated legislative response to bolster the defenses of the American technology sector. The company pointed out that illicit capability extraction campaigns undermine the investments made by AI developers and pose challenges to technology security. By raising these concerns directly with the two senators, Anthropic aims to prompt congressional action that would deter future campaigns involving the use of fraudulent accounts to systematically harvest proprietary data.
美中AI竞争与制裁关联图US-China AI Competition and Sanctions Relation Graph
展示Anthropic、阿里巴巴、美国政府及中国政府之间在AI技术出口管制与军事黑名单制裁中的核心地缘政治关系。
Illustrates the core geopolitical relationships among Anthropic, Alibaba, the US Government, and the Chinese Government regarding AI export controls and military blacklists.
04市场反应：阿里巴巴股价应声下跌Market Reaction: Alibaba Shares Decline Following Accusations
在Anthropic指控阿里巴巴对Claude模型实施大规模“蒸馏攻击”的消息曝光后，资本市场迅速做出了反应。周三下午，在纽约证券交易所交易的阿里巴巴集团美国存托凭证（ADR）价格出现明显下滑。这一指控引发了投资者对这家中国科技与电子商务巨头面临潜在监管、法律或技术合作风险的担忧，导致其股价承压。
Capital markets reacted swiftly to the news of Anthropic's allegations accusing Alibaba of conducting a massive distillation attack on its Claude AI model. On Wednesday afternoon, Alibaba Group's American depositary receipts (ADRs) experienced a notable decline in New York trading. The disclosure of the formal complaint sent to US lawmakers sparked immediate investor concern regarding potential regulatory, legal, or operational repercussions for the Chinese e-commerce and technology conglomerate.
根据交易数据显示，在纽约时间周三下午3点38分，阿里巴巴的ADR价格一路走低，触及了当天的盘中低点。其股价在下午交易时段内下跌了超过3%，最终跌至每股99.10美元。这一市场表现直接反映了投资者对两家AI巨头之间技术纠纷以及未来美国国会可能采取惩罚性措施的即时反应。
According to market data, Alibaba's ADRs fell to a session low on Wednesday afternoon, dropping more than 3% to reach $99.10 at 3:38 p.m. in New York. This downward movement directly reflects the immediate market impact of the public dispute between the two technology firms, as investors assessed the potential consequences of Anthropic's appeal for congressional intervention and penalties against the Chinese company.
