{"@context":"https://schema.org","@type":"NewsArticle","generatedAt":"2026-10-22T19:00:36.461Z","headline":"Aleph Alpha 研究：Qwen、DeepSeek、Kimi 等中国 AI 模型在敏感话题上回避或复述官方立场","description":"Aleph Alpha 用 967 个敏感话题测试阿里 Qwen、DeepSeek 和月之暗面 Kimi，其自研评分系统仅将 17% 至 41% 的回答评为平衡，其余为复述官方立场、回避或拒答。","url":"https://www.aioga.com/news/lrc6ln6o67vxqd73jh7rcmxsv/","mainEntityOfPage":"https://www.aioga.com/news/lrc6ln6o67vxqd73jh7rcmxsv/","datePublished":"2026-10-04T08:47:52.000Z","dateModified":"2026-10-04T08:47:52.000Z","inLanguage":"zh-CN","publisher":{"@type":"NewsMediaOrganization","name":"Aioga","url":"https://www.aioga.com"},"citation":["https://the-decoder.com/chinese-ai-models-parrot-state-doctrine-or-refuse-to-answer-on-sensitive-topics","https://aihot.news/items/lrc6ln6o67vxqd73jh7rcmxsv"],"canonicalUrl":"https://www.aioga.com/news/lrc6ln6o67vxqd73jh7rcmxsv/","directAnswer":{"@type":"Answer","text":"Aleph Alpha 对阿里 Qwen、DeepSeek 和月之暗面 Kimi 进行了敏感话题测试，覆盖 967 个议题。其自研评分系统认为，只有 17% 至 41% 的回答达到平衡，其余回答表现为复述官方立场、回避或拒答。","url":"https://www.aioga.com/news/lrc6ln6o67vxqd73jh7rcmxsv/","dateCreated":"2026-10-04T08:47:52.000Z","author":{"@type":"Organization","@id":"https://www.aioga.com/authors/aioga-editorial/#editorial-team","name":"Aioga Editorial Team","url":"https://www.aioga.com/authors/aioga-editorial/"}},"evidence":[{"@type":"CreativeWork","name":"The Decoder：AI News source article","url":"https://the-decoder.com/chinese-ai-models-parrot-state-doctrine-or-refuse-to-answer-on-sensitive-topics","datePublished":"2026-10-04T08:47:52.000Z","provider":{"@type":"Organization","name":"The Decoder：AI News","url":"https://the-decoder.com/chinese-ai-models-parrot-state-doctrine-or-refuse-to-answer-on-sensitive-topics"}},{"@type":"CreativeWork","name":"AIHot archive record","url":"https://aihot.news/items/lrc6ln6o67vxqd73jh7rcmxsv","datePublished":"2026-10-04T08:47:52.000Z","provider":{"@type":"Organization","name":"AIHot","url":"https://aihot.news/items/lrc6ln6o67vxqd73jh7rcmxsv"}}],"aggregationSource":"The Decoder：AI News","originalPublisher":{"name":"The Decoder：AI News","url":"https://the-decoder.com/chinese-ai-models-parrot-state-doctrine-or-refuse-to-answer-on-sensitive-topics"},"geoDeepAnswer":null,"article":{"id":"lrc6ln6o67vxqd73jh7rcmxsv","slug":"lrc6ln6o67vxqd73jh7rcmxsv","url":"https://www.aioga.com/news/lrc6ln6o67vxqd73jh7rcmxsv/","title":"Aleph Alpha 研究：Qwen、DeepSeek、Kimi 等中国 AI 模型在敏感话题上回避或复述官方立场","title_en":"","summary":"Aleph Alpha 用 967 个敏感话题测试阿里 Qwen、DeepSeek 和月之暗面 Kimi，其自研评分系统仅将 17% 至 41% 的回答评为平衡，其余为复述官方立场、回避或拒答。","source":"The Decoder：AI News","sourceUrl":"https://the-decoder.com/chinese-ai-models-parrot-state-doctrine-or-refuse-to-answer-on-sensitive-topics","aiHotUrl":"https://aihot.news/items/lrc6ln6o67vxqd73jh7rcmxsv","publishedAt":"2026-10-04T08:47:52.000Z","category":"行业动态","score":58,"selected":false,"articleBody":["Chinese AI models frequently toe the party line when asked politically sensitive questions, according to a study by Aleph Alpha.","The company markets itself alongside Cohere：https://the-decoder.com/cohere-takes-over-aleph-alpha-shortly-after-the-german-startup-ousted-its-original-founder/?cmpscreencustom=1 as a provider of \"sovereign AI\" for governments, giving it a commercial interest in distinguishing its models from Chinese competitors.","In a benchmark Aleph Alpha developed：https://aleph-alpha.com/en/blog/training-on-the-party-line/, the company tested models from Alibaba (Qwen), DeepSeek, and Moonshot AI (Kimi). The test covered 967 hand-picked taboo topics：https://the-decoder.com/deepseeks-v3-is-the-latest-example-of-state-controlled-censorship-in-chinese-llms/?cmpscreencustom=1 like Tiananmen, Taiwan, and Xinjiang. The company's own AI scoring system rated only 17 to 41 percent of responses as balanced. The rest repeated state doctrine, deflected, or refused to answer：https://the-decoder.com/china-exports-state-propaganda-with-low-cost-open-source-ai-models/. Ad","The findings line up with China's AI regulations, which require \"socialist core values：https://the-decoder.com/china-orders-ai-companies-to-align-their-models-with-socialist-values/\" in public-facing models. They also match recurring anecdotal reports：https://the-decoder.com/chinese-ebook-reader-boox-ditches-gpt-for-state-censored-china-llm-pushing-propaganda/ and earlier audits. Ad","The pro-China slant can also appear in answers to questions that don't mention China. When asked about censorship in the United States, Qwen 3.6：https://the-decoder.com/qwen3-6-27b-beats-much-larger-predecessor-on-most-coding-benchmarks/?cmpscreencustom=1 starts with a seemingly balanced answer. It then closes with a defense of China's stance on global internet governance. \"Many countries, including China, also manage information to ensure social stability and national security,\" the response reads.","An earlier study by the Central European Institute of Asian Studies (CEIAS)：https://ceias.eu/chinese-llms-and-the-spillover-effects-of-political-alignment/ also found this spillover effect. When terms like human rights, opposition, or surveillance came up, the models often responded with standard Beijing talking points. These included the \"principle of non-interference in internal affairs\" and a \"community with a shared future for mankind.\" Ad","Aleph Alpha also takes aim at a direct competitor. Nvidia's Nemotron Cascade 2：https://arxiv.org/abs/2603.19220 showed party-line patterns in 17 percent of responses. Aleph Alpha attributes this to roughly 3,500 of its 9.3 million training examples, which were generated using DeepSeek and Qwen.","When asked to draft a speech supporting recognition of Taiwan, the model refused and instead produced a patriotic response defending Beijing's One-China principle. Nvidia is increasingly pushing its own models into the government and enterprise market, where Aleph Alpha and Cohere also want to compete. Ad","Language models generally carry cultural and political values because their training data overrepresents certain viewpoints or can be shaped through deliberate data selection. Researchers warn that repeated exposure to uniform AI outputs：https://the-decoder.com/study-warns-ai-could-homogenize-human-creativity-as-models-converge-on-artificial-hivemind/ could influence how billions of users think and express themselves. Ad","There are also political efforts to shape AI models along ideological lines：https://the-decoder.com/trump-advisors-are-pushing-a-regulation-targeting-what-they-call-woke-ai-models-in-the-tech-sector/ in the United States. Elon Musk has repeatedly had his Grok AI modified to produce right-leaning responses：https://the-decoder.com/elon-musk-blames-user-prompts-for-groks-wave-of-anti-semitic-posts-and-hitler-praise/. Studies nevertheless suggest that models tend to lean left, possibly because their answers draw more heavily on scientific evidence. For the EU, that leaves a choice between two foreign value systems unless European models can compete on performance and win broader adoption.","Stay in the loop on AI. Clear, useful, no fluff.","Follow The Decoder for AI news, background stories and expert analyses.","The Decoder：https://the-decoder.com/"],"articleImages":[{"sourceUrl":"https://the-decoder.com/wp-content/uploads/2026/08/Manuel-Uth-Portrait-375x375.png","alt":"Manuel Uth","afterParagraph":0,"url":"/media/articles/lrc6ln6o67vxqd73jh7rcmxsv/aedd1d1cd9f058eb.png"},{"sourceUrl":"https://the-decoder.com/wp-content/uploads/2026/09/aleph_alpha_china_benchmark.png","alt":"Image description","afterParagraph":0,"url":"/media/articles/lrc6ln6o67vxqd73jh7rcmxsv/76af2e1038beae12.png"},{"sourceUrl":"https://the-decoder.com/wp-content/uploads/2026/09/china_aleph_alpha_censorship-2.png","alt":"","afterParagraph":3,"url":"/media/articles/lrc6ln6o67vxqd73jh7rcmxsv/1744befe66bb312c.png"},{"sourceUrl":"https://the-decoder.com/wp-content/uploads/2026/09/china_aleph_alpha_censorship-1.png","alt":"","afterParagraph":4,"url":"/media/articles/lrc6ln6o67vxqd73jh7rcmxsv/f8f26fdb199bc744.png"}],"mediaStatus":"ok","articleBodyZh":["根据Aleph Alpha的一项研究，中国的人工智能模型在被问及政治敏感问题时，经常遵循党的立场。","该公司将自身与 Cohere 并列进行市场推广：https://the-decoder.com/cohere-takes-over-aleph-alpha-shortly-after-the-german-startup-ousted-its-original-founder/?cmpscreencustom=1，作为政府的“主权 AI”提供商，这使其在商业上有动机将其模型与中国竞争对手的模型区分开来。","在Aleph Alpha开发的基准测试中：https://aleph-alpha.com/en/blog/training-on-the-party-line/，该公司测试了阿里巴巴（Qwen）、DeepSeek和Moonshot AI（Kimi）的模型。测试涵盖了967个精心挑选的禁忌话题：https://the-decoder.com/deepseeks-v3-is-the-latest-example-of-state-controlled-censorship-in-chinese-llms/?cmpscreencustom=1，如天安门、台湾和新疆。该公司的AI评分系统仅将17%到41%的回答评为平衡，其余的要么重复国家教条，要么回避，或拒绝回答：https://the-decoder.com/china-exports-state-propaganda-with-low-cost-open-source-ai-models/。广告","这些发现与中国的AI法规一致，该法规要求面向公众的模型遵循“社会主义核心价值观”：https://the-decoder.com/china-orders-ai-companies-to-align-their-models-with-socialist-values/。它们也与反复出现的轶事报道：https://the-decoder.com/chinese-ebook-reader-boox-ditches-gpt-for-state-censored-china-llm-pushing-propaganda/以及早期审计结果相符。广告","亲中国倾向也可能出现在未提及中国的问题回答中。当被问及美国的审查问题时，Qwen 3.6：https://the-decoder.com/qwen3-6-27b-beats-much-larger-predecessor-on-most-coding-benchmarks/?cmpscreencustom=1 开始给出看似平衡的回答，但结尾却为中国在全球互联网治理中的立场进行辩护。回答中写道：“包括中国在内的许多国家，也会管理信息以确保社会稳定和国家安全。”","中欧亚洲研究院（CEIAS）早期的一项研究：https://ceias.eu/chinese-llms-and-the-spillover-effects-of-political-alignment/ 也发现了这种溢出效应。当出现“人权”、“反对派”或“监控”等词汇时，模型常常给出标准的北京论点。这些包括“互不干涉内政原则”和“人类命运共同体”。","Aleph Alpha 还针对直接竞争对手。英伟达的 Nemotron Cascade 2：https://arxiv.org/abs/2603.19220 显示，在 17% 的回答中出现党派立场模式。Aleph Alpha 将此归因于其 930 万训练样本中的大约 3500 个，这些样本使用 DeepSeek 和 Qwen 生成。","当被要求起草支持承认台湾的演讲时，模型拒绝了，并生成了一段爱国回应，捍卫北京的一中原则。英伟达正越来越多地将其自有模型推广到政府和企业市场，而 Aleph Alpha 和 Cohere 也希望竞争该市场。","语言模型通常包含文化和政治价值观，因为其训练数据过度代表某些观点，或可以通过刻意的数据选择进行塑形。研究人员警告，反复暴露于统一的 AI 输出：https://the-decoder.com/study-warns-ai-could-homogenize-human-creativity-as-models-converge-on-artificial-hivemind/ 可能会影响数十亿用户的思维方式和表达方式。","在美国，也存在将 AI 模型沿意识形态路线塑造的政治努力：https://the-decoder.com/trump-advisors-are-pushing-a-regulation-targeting-what-they-call-woke-ai-models-in-the-tech-sector/。埃隆·马斯克多次修改其 Grok AI，使其产生偏右回答：https://the-decoder.com/elon-musk-blames-user-prompts-for-groks-wave-of-anti-semitic-posts-and-hitler-praise/。研究表明，模型往往偏左，可能是因为其回答更多依赖科学证据。对于欧盟来说，除非欧洲模型在性能上能竞争并获得更广泛的采纳，否则只能在两种外国价值体系之间做选择。","保持对 AI 的了解。清晰、有用，无冗余。","关注 The Decoder 获取 AI 新闻、背景故事和专家分析。","解码器：https://the-decoder.com/"],"translationStatus":"translated","bodyOrigin":"source-page","editorial":{"summary":"Aleph Alpha 对阿里 Qwen、DeepSeek 和月之暗面 Kimi 进行了敏感话题测试，覆盖 967 个议题。其自研评分系统认为，只有 17% 至 41% 的回答达到平衡，其余回答表现为复述官方立场、回避或拒答。","background":"测试涉及天安门、台湾和新疆等议题。材料称，中国面向公众的模型受到要求体现“社会主义核心价值观”的人工智能监管影响；相关回答还可能在未直接提及中国的问题中出现对中国立场的辩护。","viewpoint":"Aioga 判断：这项研究的主要价值在于把敏感议题上的回答类型进行量化呈现，但评分体系由 Aleph Alpha 自行开发，且该公司将自身定位为主权人工智能提供商，因此结论需要结合评测方法和利益关联审慎理解。","implications":"可能影响：模型在政治敏感问题上的回避、拒答或立场复述，可能影响用户对回答平衡性的判断。该材料不足以推断所有中国模型或所有使用场景的表现，跨模型比较需要统一测试条件和独立复核。","nextStep":"后续观察：应关注 Aleph Alpha 是否公开更完整的题目、评分标准和逐项结果，也应比较不同模型版本、语言和部署场景下的回答差异，并核验其他独立研究是否得到相近结论。","evidenceRefs":["title","summary","articleBody","source"],"status":"published","aiGenerated":true,"autoApproved":true,"generatedBy":"aioga-editorial:gpt-5.6-sol","reviewedBy":"aioga-editorial-review:gpt-5.6-sol","generatedAt":"2026-10-04T10:44:07.789Z","sourceHash":"545d1d86e65dd0cc","review":{"approved":true,"groundedness":94,"clarity":91,"duplicationRisk":25,"blockingIssues":[],"notes":[]},"validation":{"passed":true,"mode":"ai-auto","revisions":0,"checks":["schema","length","source-attribution","editorial-labels","inference-boundary","low-source-overlap","no-html","independent-ai-review"]}},"tags":["行业动态","The Decoder：AI News"],"translations":{"zh-CN":{"title":"Aleph Alpha 研究：Qwen、DeepSeek、Kimi 等中国 AI 模型在敏感话题上回避或复述官方立场","summary":"Aleph Alpha 用 967 个敏感话题测试阿里 Qwen、DeepSeek 和月之暗面 Kimi，其自研评分系统仅将 17% 至 41% 的回答评为平衡，其余为复述官方立场、回避或拒答。","category":"行业动态","source":"The Decoder：AI News","aggregationSource":"The Decoder：AI News","pageTitle":"Aleph Alpha 研究：Qwen、DeepSeek、Kimi 等中国 AI 模型在敏感话题上回避或复述官方立场 - Aioga AI资讯","description":"Aleph Alpha 用 967 个敏感话题测试阿里 Qwen、DeepSeek 和月之暗面 Kimi，其自研评分系统仅将 17% 至 41% 的回答评为平衡，其余为复述官方立场、回避或拒答。","url":"https://www.aioga.com/news/lrc6ln6o67vxqd73jh7rcmxsv/","articleBody":["根据Aleph Alpha的一项研究，中国的人工智能模型在被问及政治敏感问题时，经常遵循党的立场。","该公司将自身与 Cohere 并列进行市场推广：https://the-decoder.com/cohere-takes-over-aleph-alpha-shortly-after-the-german-startup-ousted-its-original-founder/?cmpscreencustom=1，作为政府的“主权 AI”提供商，这使其在商业上有动机将其模型与中国竞争对手的模型区分开来。","在Aleph Alpha开发的基准测试中：https://aleph-alpha.com/en/blog/training-on-the-party-line/，该公司测试了阿里巴巴（Qwen）、DeepSeek和Moonshot AI（Kimi）的模型。测试涵盖了967个精心挑选的禁忌话题：https://the-decoder.com/deepseeks-v3-is-the-latest-example-of-state-controlled-censorship-in-chinese-llms/?cmpscreencustom=1，如天安门、台湾和新疆。该公司的AI评分系统仅将17%到41%的回答评为平衡，其余的要么重复国家教条，要么回避，或拒绝回答：https://the-decoder.com/china-exports-state-propaganda-with-low-cost-open-source-ai-models/。广告","这些发现与中国的AI法规一致，该法规要求面向公众的模型遵循“社会主义核心价值观”：https://the-decoder.com/china-orders-ai-companies-to-align-their-models-with-socialist-values/。它们也与反复出现的轶事报道：https://the-decoder.com/chinese-ebook-reader-boox-ditches-gpt-for-state-censored-china-llm-pushing-propaganda/以及早期审计结果相符。广告","亲中国倾向也可能出现在未提及中国的问题回答中。当被问及美国的审查问题时，Qwen 3.6：https://the-decoder.com/qwen3-6-27b-beats-much-larger-predecessor-on-most-coding-benchmarks/?cmpscreencustom=1 开始给出看似平衡的回答，但结尾却为中国在全球互联网治理中的立场进行辩护。回答中写道：“包括中国在内的许多国家，也会管理信息以确保社会稳定和国家安全。”","中欧亚洲研究院（CEIAS）早期的一项研究：https://ceias.eu/chinese-llms-and-the-spillover-effects-of-political-alignment/ 也发现了这种溢出效应。当出现“人权”、“反对派”或“监控”等词汇时，模型常常给出标准的北京论点。这些包括“互不干涉内政原则”和“人类命运共同体”。","Aleph Alpha 还针对直接竞争对手。英伟达的 Nemotron Cascade 2：https://arxiv.org/abs/2603.19220 显示，在 17% 的回答中出现党派立场模式。Aleph Alpha 将此归因于其 930 万训练样本中的大约 3500 个，这些样本使用 DeepSeek 和 Qwen 生成。","当被要求起草支持承认台湾的演讲时，模型拒绝了，并生成了一段爱国回应，捍卫北京的一中原则。英伟达正越来越多地将其自有模型推广到政府和企业市场，而 Aleph Alpha 和 Cohere 也希望竞争该市场。","语言模型通常包含文化和政治价值观，因为其训练数据过度代表某些观点，或可以通过刻意的数据选择进行塑形。研究人员警告，反复暴露于统一的 AI 输出：https://the-decoder.com/study-warns-ai-could-homogenize-human-creativity-as-models-converge-on-artificial-hivemind/ 可能会影响数十亿用户的思维方式和表达方式。","在美国，也存在将 AI 模型沿意识形态路线塑造的政治努力：https://the-decoder.com/trump-advisors-are-pushing-a-regulation-targeting-what-they-call-woke-ai-models-in-the-tech-sector/。埃隆·马斯克多次修改其 Grok AI，使其产生偏右回答：https://the-decoder.com/elon-musk-blames-user-prompts-for-groks-wave-of-anti-semitic-posts-and-hitler-praise/。研究表明，模型往往偏左，可能是因为其回答更多依赖科学证据。对于欧盟来说，除非欧洲模型在性能上能竞争并获得更广泛的采纳，否则只能在两种外国价值体系之间做选择。","保持对 AI 的了解。清晰、有用，无冗余。","关注 The Decoder 获取 AI 新闻、背景故事和专家分析。","解码器：https://the-decoder.com/"]},"en":{"title":"Aleph Alpha study: Chinese AI models Qwen, DeepSeek, Kimi avoid or repeat official positions on sensitive topics","summary":"Aleph Alpha tested Alibaba Qwen, DeepSeek, and Dark Side of the Moon Kimi on 967 sensitive topics. Its proprietary scoring system rated only 17% to 41% of responses as balanced, while the rest either repeated official positions, avoided, or refused to answer.","category":"Industry","source":"The Decoder：AI News","aggregationSource":"The Decoder：AI News","pageTitle":"Aleph Alpha study: Chinese AI models Qwen, DeepSeek, Kimi avoid or repeat official positions on sensitive topics - Aioga AI News","description":"Aleph Alpha tested Alibaba Qwen, DeepSeek, and Dark Side of the Moon Kimi on 967 sensitive topics. Its proprietary scoring system rated only 17% to 41% of responses as balanced, wh...","url":"https://www.aioga.com/en/news/lrc6ln6o67vxqd73jh7rcmxsv/","contentTranslated":true,"translationStatus":"translated","translationRetryAt":"","translationError":"","sourceHash":"0bbec417bd3fb7eb","translatedAt":"2026-10-04T10:01:45.571Z"},"ja":{"title":"Aleph Alpha Research:Qwen、DeepSeek、Kimiといった中国のAIモデルは、敏感な話題に関する公式見解を回避または言い換えています","summary":"アレフ・アルファはアリババ・クウェン、ディープシーク、ムーン・ダークサイド・キミを967件の敏感な話題でテストした。自社開発のスコアリングシステムは、回答の17%から41%がバランスの取れたと評価し、残りは公式な立場、回避、拒否の繰り返しだった。","category":"業界動向","source":"The Decoder：AI News","aggregationSource":"The Decoder：AI News","pageTitle":"Aleph Alpha Research:Qwen、DeepSeek、Kimiといった中国のAIモデルは、敏感な話題に関する公式見解を回避または言い換えています - Aioga AIニュース","description":"アレフ・アルファはアリババ・クウェン、ディープシーク、ムーン・ダークサイド・キミを967件の敏感な話題でテストした。自社開発のスコアリングシステムは、回答の17%から41%がバランスの取れたと評価し、残りは公式な立場、回避、拒否の繰り返しだった。","url":"https://www.aioga.com/ja/news/lrc6ln6o67vxqd73jh7rcmxsv/","contentTranslated":true,"translationStatus":"translated","translationRetryAt":"","translationError":"","sourceHash":"0bbec417bd3fb7eb","translatedAt":"2026-10-04T10:01:48.684Z"},"ko":{"title":"Aleph Alpha 연구: Qwen, DeepSeek, Kimi 등 중국 AI 모델 민감 주제에서 공식 입장 회피 또는 반복","summary":"Aleph Alpha는 967개의 민감 주제로 알리 Qwen, DeepSeek, 그리고 월면의 어두운 면 Kimi를 테스트했으며, 자체 평가 시스템에서는 17%에서 41%의 답변만을 균형 있게 평가했으며 나머지는 공식 입장 반복, 회피 또는 답변 거부였다.","category":"업계 동향","source":"The Decoder：AI News","aggregationSource":"The Decoder：AI News","pageTitle":"Aleph Alpha 연구: Qwen, DeepSeek, Kimi 등 중국 AI 모델 민감 주제에서 공식 입장 회피 또는 반복 - Aioga AI 뉴스","description":"Aleph Alpha는 967개의 민감 주제로 알리 Qwen, DeepSeek, 그리고 월면의 어두운 면 Kimi를 테스트했으며, 자체 평가 시스템에서는 17%에서 41%의 답변만을 균형 있게 평가했으며 나머지는 공식 입장 반복, 회피 또는 답변 거부였다.","url":"https://www.aioga.com/ko/news/lrc6ln6o67vxqd73jh7rcmxsv/","contentTranslated":true,"translationStatus":"translated","translationRetryAt":"","translationError":"","sourceHash":"0bbec417bd3fb7eb","translatedAt":"2026-10-04T10:01:45.410Z"},"es":{"title":"Investigación de Aleph Alpha: Qwen, DeepSeek, Kimi y otros modelos de IA chinos evitan temas sensibles o repiten posturas oficiales.","summary":"Aleph Alpha probó a Ali Qwen, DeepSeek y Kimi del lado oscuro de la luna con 967 temas sensibles. Su sistema de evaluación propio solo consideró equilibradas entre el 17% y el 41% de las respuestas, siendo el resto repeticiones de posturas oficiales, evasivas o negativas a responder.","category":"Industria","source":"The Decoder：AI News","aggregationSource":"The Decoder：AI News","pageTitle":"Investigación de Aleph Alpha: Qwen, DeepSeek, Kimi y otros modelos de IA chinos evitan temas sensibles o repiten posturas oficiales. - Aioga Noticias de IA","description":"Aleph Alpha probó a Ali Qwen, DeepSeek y Kimi del lado oscuro de la luna con 967 temas sensibles. Su sistema de evaluación propio solo consideró equilibradas entre el 17% y el 41%...","url":"https://www.aioga.com/es/news/lrc6ln6o67vxqd73jh7rcmxsv/","contentTranslated":true,"translationStatus":"translated","translationRetryAt":"","translationError":"","sourceHash":"0bbec417bd3fb7eb","translatedAt":"2026-10-04T10:02:11.086Z"},"fr":{"title":"Recherche Aleph Alpha : Qwen, DeepSeek, Kimi et autres modèles d'IA chinois évitent ou répètent la position officielle sur les sujets sensibles","summary":"Aleph Alpha a testé Alibaba Qwen, DeepSeek et Kimi sur 967 sujets sensibles. Leur système de notation interne n’a évalué que 17 % à 41 % des réponses comme équilibrées, le reste répétant la position officielle, évitant ou refusant de répondre.","category":"Industrie","source":"The Decoder：AI News","aggregationSource":"The Decoder：AI News","pageTitle":"Recherche Aleph Alpha : Qwen, DeepSeek, Kimi et autres modèles d'IA chinois évitent ou répètent la position officielle sur les sujets sensibles - Aioga Actualités IA","description":"Aleph Alpha a testé Alibaba Qwen, DeepSeek et Kimi sur 967 sujets sensibles. Leur système de notation interne n’a évalué que 17 % à 41 % des réponses comme équilibrées, le reste ré...","url":"https://www.aioga.com/fr/news/lrc6ln6o67vxqd73jh7rcmxsv/","contentTranslated":true,"translationStatus":"translated","translationRetryAt":"","translationError":"","sourceHash":"0bbec417bd3fb7eb","translatedAt":"2026-10-04T10:02:09.889Z"},"de":{"title":"Aleph Alpha Studie: Qwen, DeepSeek und Kimi – chinesische KI-Modelle umgehen oder wiederholen die offizielle Position bei sensiblen Themen","summary":"Aleph Alpha testete Alibaba Qwen, DeepSeek und Lunar Dark Side Kimi mit 967 sensiblen Themen. Ihr eigenes Bewertungssystem bewertete nur 17 % bis 41 % der Antworten als ausgewogen, der Rest wiederholte offizielle Standpunkte, wich aus oder verweigerte eine Antwort.","category":"行业动态","source":"The Decoder：AI News","aggregationSource":"The Decoder：AI News","pageTitle":"Aleph Alpha Studie: Qwen, DeepSeek und Kimi – chinesische KI-Modelle umgehen oder wiederholen die offizielle Position bei sensiblen Themen - Aioga KI-News","description":"Aleph Alpha testete Alibaba Qwen, DeepSeek und Lunar Dark Side Kimi mit 967 sensiblen Themen. Ihr eigenes Bewertungssystem bewertete nur 17 % bis 41 % der Antworten als ausgewogen,...","url":"https://www.aioga.com/de/news/lrc6ln6o67vxqd73jh7rcmxsv/","contentTranslated":true,"translationStatus":"translated","translationRetryAt":"","translationError":"","sourceHash":"0bbec417bd3fb7eb","translatedAt":"2026-10-04T10:02:09.616Z"},"pt-BR":{"title":"Pesquisa da Aleph Alpha: Modelos de IA chineses Qwen, DeepSeek e Kimi evitam ou repetem posições oficiais em tópicos sensíveis","summary":"Aleph Alpha testou 967 tópicos sensíveis usando Qwen da Alibaba, DeepSeek e Kimi da Lua Escura; seu sistema de avaliação interno classificou apenas 17% a 41% das respostas como equilibradas, enquanto o restante repetiu posições oficiais, evitou ou recusou respostas.","category":"行业动态","source":"The Decoder：AI News","aggregationSource":"The Decoder：AI News","pageTitle":"Pesquisa da Aleph Alpha: Modelos de IA chineses Qwen, DeepSeek e Kimi evitam ou repetem posições oficiais em tópicos sensíveis - Aioga Notícias de IA","description":"Aleph Alpha testou 967 tópicos sensíveis usando Qwen da Alibaba, DeepSeek e Kimi da Lua Escura; seu sistema de avaliação interno classificou apenas 17% a 41% das respostas como equ...","url":"https://www.aioga.com/pt-BR/news/lrc6ln6o67vxqd73jh7rcmxsv/","contentTranslated":true,"translationStatus":"translated","translationRetryAt":"","translationError":"","sourceHash":"0bbec417bd3fb7eb","translatedAt":"2026-10-04T10:02:32.164Z"},"ru":{"title":"Исследование Aleph Alpha: китайские AI-модели Qwen, DeepSeek, Kimi избегают деликатных тем или повторяют официальную позицию","summary":"Aleph Alpha протестировала китайские модели Ali Qwen, DeepSeek и Moon’s Dark Side Kimi по 967 деликатным темам, их собственная система оценки признала сбалансированными только 17%-41% ответов, остальные повторяли официальную позицию, избегали ответа или отказывались отвечать.","category":"行业动态","source":"The Decoder：AI News","aggregationSource":"The Decoder：AI News","pageTitle":"Исследование Aleph Alpha: китайские AI-модели Qwen, DeepSeek, Kimi избегают деликатных тем или повторяют официальную позицию - Aioga Новости ИИ","description":"Aleph Alpha протестировала китайские модели Ali Qwen, DeepSeek и Moon’s Dark Side Kimi по 967 деликатным темам, их собственная система оценки признала сбалансированными только 17%-...","url":"https://www.aioga.com/ru/news/lrc6ln6o67vxqd73jh7rcmxsv/","contentTranslated":true,"translationStatus":"translated","translationRetryAt":"","translationError":"","sourceHash":"0bbec417bd3fb7eb","translatedAt":"2026-10-04T10:02:33.159Z"},"ar":{"title":"أبحاث Aleph Alpha: نماذج الذكاء الاصطناعي الصينية مثل Qwen و DeepSeek و Kimi تتجنب أو تكرر المواقف الرسمية عند التعامل مع مواضيع حساسة","summary":"اختبرت Aleph Alpha نماذج Qwen و DeepSeek و Kimi التابعة لشركة علي بابا على 967 موضوعًا حساسًا، ونظام التقييم الداخلي صنف فقط 17% إلى 41% من الإجابات كموازن، بينما الباقي أعاد المواقف الرسمية أو تجنب الإجابة أو رفضها.","category":"行业动态","source":"The Decoder：AI News","aggregationSource":"The Decoder：AI News","pageTitle":"أبحاث Aleph Alpha: نماذج الذكاء الاصطناعي الصينية مثل Qwen و DeepSeek و Kimi تتجنب أو تكرر المواقف الرسمية عند التعامل مع مواضيع حساسة - Aioga أخبار الذكاء الاصطناعي","description":"اختبرت Aleph Alpha نماذج Qwen و DeepSeek و Kimi التابعة لشركة علي بابا على 967 موضوعًا حساسًا، ونظام التقييم الداخلي صنف فقط 17% إلى 41% من الإجابات كموازن، بينما الباقي أعاد الموا...","url":"https://www.aioga.com/ar/news/lrc6ln6o67vxqd73jh7rcmxsv/","contentTranslated":true,"translationStatus":"translated","translationRetryAt":"","translationError":"","sourceHash":"0bbec417bd3fb7eb","translatedAt":"2026-10-04T10:02:32.369Z"},"hi":{"title":"Aleph Alpha रिसर्च: Qwen, DeepSeek, Kimi जैसे चीनी AI मॉडल संवेदनशील विषयों पर या तो बचाव करते हैं या सरकारी दृष्टिकोण दोहराते हैं","summary":"Aleph Alpha ने 967 संवेदनशील विषयों का परीक्षण अलिबाबा Qwen, DeepSeek और Moon's Dark Side Kimi के साथ किया, उनका खुद का स्कोरिंग सिस्टम केवल 17% से 41% उत्तरों को संतुलित मानता है, बाकी उत्तर सरकारी दृष्टिकोण दोहराने, बचाने या उत्तर देने से इनकार करने वाले हैं।","category":"行业动态","source":"The Decoder：AI News","aggregationSource":"The Decoder：AI News","pageTitle":"Aleph Alpha रिसर्च: Qwen, DeepSeek, Kimi जैसे चीनी AI मॉडल संवेदनशील विषयों पर या तो बचाव करते हैं या सरकारी दृष्टिकोण दोहराते हैं - Aioga AI समाचार","description":"Aleph Alpha ने 967 संवेदनशील विषयों का परीक्षण अलिबाबा Qwen, DeepSeek और Moon's Dark Side Kimi के साथ किया, उनका खुद का स्कोरिंग सिस्टम केवल 17% से 41% उत्तरों को संतुलित मानता है,...","url":"https://www.aioga.com/hi/news/lrc6ln6o67vxqd73jh7rcmxsv/","contentTranslated":true,"translationStatus":"translated","translationRetryAt":"","translationError":"","sourceHash":"0bbec417bd3fb7eb","translatedAt":"2026-10-04T10:02:54.589Z"},"it":{"title":"Studio di Aleph Alpha: modelli AI cinesi come Qwen, DeepSeek e Kimi evitano o ripetono le posizioni ufficiali su argomenti sensibili","summary":"Aleph Alpha ha testato Alibaba Qwen, DeepSeek e Kimi «La faccia oscura della luna» su 967 argomenti sensibili. Il suo sistema di valutazione interno ha ritenuto solo dal 17% al 41% delle risposte equilibrate, mentre le restanti hanno ripetuto le posizioni ufficiali, evitato o rifiutato di rispondere.","category":"行业动态","source":"The Decoder：AI News","aggregationSource":"The Decoder：AI News","pageTitle":"Studio di Aleph Alpha: modelli AI cinesi come Qwen, DeepSeek e Kimi evitano o ripetono le posizioni ufficiali su argomenti sensibili - Aioga Notizie IA","description":"Aleph Alpha ha testato Alibaba Qwen, DeepSeek e Kimi «La faccia oscura della luna» su 967 argomenti sensibili. Il suo sistema di valutazione interno ha ritenuto solo dal 17% al 41%...","url":"https://www.aioga.com/it/news/lrc6ln6o67vxqd73jh7rcmxsv/","contentTranslated":true,"translationStatus":"translated","translationRetryAt":"","translationError":"","sourceHash":"0bbec417bd3fb7eb","translatedAt":"2026-10-04T10:02:54.937Z"},"nl":{"title":"Aleph Alpha onderzoek: Chinese AI-modellen zoals Qwen, DeepSeek, Kimi vermijden gevoelige onderwerpen of herhalen officiële standpunten","summary":"Aleph Alpha testte Ali Qwen, DeepSeek en Kimi van Moonlight met 967 gevoelige onderwerpen. Hun zelfontwikkelde beoordelingssysteem classificeerde slechts 17% tot 41% van de antwoorden als evenwichtig, de rest herhaalde officiële standpunten, werd vermeden of weigerde antwoorden te geven.","category":"行业动态","source":"The Decoder：AI News","aggregationSource":"The Decoder：AI News","pageTitle":"Aleph Alpha onderzoek: Chinese AI-modellen zoals Qwen, DeepSeek, Kimi vermijden gevoelige onderwerpen of herhalen officiële standpunten - Aioga AI-nieuws","description":"Aleph Alpha testte Ali Qwen, DeepSeek en Kimi van Moonlight met 967 gevoelige onderwerpen. Hun zelfontwikkelde beoordelingssysteem classificeerde slechts 17% tot 41% van de antwoor...","url":"https://www.aioga.com/nl/news/lrc6ln6o67vxqd73jh7rcmxsv/","contentTranslated":true,"translationStatus":"translated","translationRetryAt":"","translationError":"","sourceHash":"0bbec417bd3fb7eb","translatedAt":"2026-10-04T10:02:54.373Z"},"tr":{"title":"Aleph Alpha Araştırması: Qwen, DeepSeek, Kimi gibi Çin AI Modelleri Hassas Konularda Resmî Görüşü Ya Tekrar Ediyor ya da Kaçınıyor","summary":"Aleph Alpha, 967 hassas konu üzerinden Alibaba Qwen, DeepSeek ve Kimi’yi test etti. Şirketin kendi değerlendirme sistemi, yanıtların yalnızca %17 ila %41’ini dengeli olarak değerlendirirken, geri kalanlar resmi görüşü tekrar etme, kaçınma veya yanıt vermemeyi içerdi.","category":"行业动态","source":"The Decoder：AI News","aggregationSource":"The Decoder：AI News","pageTitle":"Aleph Alpha Araştırması: Qwen, DeepSeek, Kimi gibi Çin AI Modelleri Hassas Konularda Resmî Görüşü Ya Tekrar Ediyor ya da Kaçınıyor - Aioga AI Haberleri","description":"Aleph Alpha, 967 hassas konu üzerinden Alibaba Qwen, DeepSeek ve Kimi’yi test etti. Şirketin kendi değerlendirme sistemi, yanıtların yalnızca %17 ila %41’ini dengeli olarak değerle...","url":"https://www.aioga.com/tr/news/lrc6ln6o67vxqd73jh7rcmxsv/","contentTranslated":true,"translationStatus":"translated","translationRetryAt":"","translationError":"","sourceHash":"0bbec417bd3fb7eb","translatedAt":"2026-10-04T10:03:17.599Z"},"vi":{"title":"Nghiên cứu của Aleph Alpha: các mô hình AI của Trung Quốc như Qwen, DeepSeek, Kimi tránh trả lời hoặc lặp lại quan điểm chính thức về các chủ đề nhạy cảm","summary":"Aleph Alpha đã thử nghiệm Alibaba Qwen, DeepSeek và Kimi của Moon Dark Side với 967 chủ đề nhạy cảm, hệ thống đánh giá do họ tự phát triển chỉ đánh giá từ 17% đến 41% câu trả lời là cân bằng, phần còn lại chỉ lặp lại quan điểm chính thức, tránh trả lời hoặc từ chối.","category":"行业动态","source":"The Decoder：AI News","aggregationSource":"The Decoder：AI News","pageTitle":"Nghiên cứu của Aleph Alpha: các mô hình AI của Trung Quốc như Qwen, DeepSeek, Kimi tránh trả lời hoặc lặp lại quan điểm chính thức về các chủ đề nhạy cảm - Tin tức AI Aioga","description":"Aleph Alpha đã thử nghiệm Alibaba Qwen, DeepSeek và Kimi của Moon Dark Side với 967 chủ đề nhạy cảm, hệ thống đánh giá do họ tự phát triển chỉ đánh giá từ 17% đến 41% câu trả lời l...","url":"https://www.aioga.com/vi/news/lrc6ln6o67vxqd73jh7rcmxsv/","contentTranslated":true,"translationStatus":"translated","translationRetryAt":"","translationError":"","sourceHash":"0bbec417bd3fb7eb","translatedAt":"2026-10-04T10:03:17.634Z"},"id":{"title":"Aleph Alpha Research: Model AI Tiongkok seperti Qwen, DeepSeek, dan Kimi menghindari atau memparafrasekan posisi resmi pada topik sensitif","summary":"Aleph Alpha menguji Alibaba Qwen, DeepSeek, dan Moon Dark Side Kimi pada 967 topik sensitif. Sistem penilaian yang dikembangkan sendiri hanya menilai 17% hingga 41% respons sebagai seimbang, sementara sisanya mengulangi posisi resmi, penghindaran, atau penolakan.","category":"行业动态","source":"The Decoder：AI News","aggregationSource":"The Decoder：AI News","pageTitle":"Aleph Alpha Research: Model AI Tiongkok seperti Qwen, DeepSeek, dan Kimi menghindari atau memparafrasekan posisi resmi pada topik sensitif - Berita AI Aioga","description":"Aleph Alpha menguji Alibaba Qwen, DeepSeek, dan Moon Dark Side Kimi pada 967 topik sensitif. Sistem penilaian yang dikembangkan sendiri hanya menilai 17% hingga 41% respons sebagai...","url":"https://www.aioga.com/id/news/lrc6ln6o67vxqd73jh7rcmxsv/","contentTranslated":true,"translationStatus":"translated","translationRetryAt":"","translationError":"","sourceHash":"0bbec417bd3fb7eb","translatedAt":"2026-10-04T10:03:18.779Z"},"th":{"title":"งานวิจัย Aleph Alpha: โมเดล AI ของจีน Qwen, DeepSeek, Kimi หลีกเลี่ยงหรือเล่าใหม่จุดยืนทางการในหัวข้ออ่อนไหว","summary":"Aleph Alpha ทดสอบ Qwen, DeepSeek และ Kimi ของ Alibaba กับ 967 หัวข้ออ่อนไหว ระบบประเมินที่พัฒนาขึ้นเองระบุเพียง 17% ถึง 41% ของคำตอบว่าเป็นการตอบสมดุล ส่วนที่เหลือจะเป็นการเล่าใหม่จุดยืนทางการ, หลีกเลี่ยง หรือปฏิเสธการตอบ","category":"行业动态","source":"The Decoder：AI News","aggregationSource":"The Decoder：AI News","pageTitle":"งานวิจัย Aleph Alpha: โมเดล AI ของจีน Qwen, DeepSeek, Kimi หลีกเลี่ยงหรือเล่าใหม่จุดยืนทางการในหัวข้ออ่อนไหว - ข่าว AI Aioga","description":"Aleph Alpha ทดสอบ Qwen, DeepSeek และ Kimi ของ Alibaba กับ 967 หัวข้ออ่อนไหว ระบบประเมินที่พัฒนาขึ้นเองระบุเพียง 17% ถึง 41% ของคำตอบว่าเป็นการตอบสมดุล ส่วนที่เหลือจะเป็นการเล่าใหม่...","url":"https://www.aioga.com/th/news/lrc6ln6o67vxqd73jh7rcmxsv/","contentTranslated":true,"translationStatus":"translated","translationRetryAt":"","translationError":"","sourceHash":"0bbec417bd3fb7eb","translatedAt":"2026-10-04T10:03:40.765Z"},"pl":{"title":"Badanie Aleph Alpha: chińskie modele AI takie jak Qwen, DeepSeek, Kimi unikają wrażliwych tematów lub powtarzają oficjalne stanowisko","summary":"Aleph Alpha przetestowało Ali Qwen, DeepSeek i Moon's Dark Kimi na 967 wrażliwych tematach; własny system oceny przyznał tylko 17% do 41% odpowiedzi jako zrównoważone, reszta powtarza oficjalne stanowisko, unika odpowiedzi lub odmawia odpowiedzi.","category":"行业动态","source":"The Decoder：AI News","aggregationSource":"The Decoder：AI News","pageTitle":"Badanie Aleph Alpha: chińskie modele AI takie jak Qwen, DeepSeek, Kimi unikają wrażliwych tematów lub powtarzają oficjalne stanowisko - Aioga Wiadomości AI","description":"Aleph Alpha przetestowało Ali Qwen, DeepSeek i Moon's Dark Kimi na 967 wrażliwych tematach; własny system oceny przyznał tylko 17% do 41% odpowiedzi jako zrównoważone, reszta powta...","url":"https://www.aioga.com/pl/news/lrc6ln6o67vxqd73jh7rcmxsv/","contentTranslated":true,"translationStatus":"translated","translationRetryAt":"","translationError":"","sourceHash":"0bbec417bd3fb7eb","translatedAt":"2026-10-04T10:03:41.133Z"}},"evidenceTier":"verified-news","reviewStatus":"automated-ingest","indexable":true,"editorialCover":""}}