[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"news-slug-anthropic-claude-invisible-watermark-eu-ai-act":3,"news-related-9f566c9a-4c39-427c-af5e-c3a6b162ec25":41},{"id":4,"title":5,"summary":6,"content":7,"original_url":8,"source_id":9,"tags":10,"translations":27,"news_slug":34,"published_at":35,"created_at":36,"modified_at":37,"is_published":38,"publish_type":39,"image_url":14,"view_count":40},"9f566c9a-4c39-427c-af5e-c3a6b162ec25","Anthropic 把不可见水印写进 Claude 文本：复制粘贴都带走的 AI 身份证","Anthropic 宣布自 2026 年 8 月 2 日起,在欧盟上线或更新的 Claude 模型生成的文本中嵌入机器可读、用户不可见的水印,并对支持的图像\u002FSVG 文件附加符合 C2PA 标准的数字签名出处信息,适用于全球范围。该举措源于 Anthropic 签署的欧盟 AI 法案第 50(2)条《AI 生成内容透明度实践守则》,旨在让平台和第三方能在文本被复制粘贴后仍可识别其由 Claude 处理过。","## 事件\n\nAnthropic 8 月 10 日通过官方支持文档确认,自 **2026 年 8 月 2 日起**在欧盟上线或此后更新的 Claude 模型,从发布第一天起就内置机器可读的内容标记机制。文本走「嵌入式水印」路线,文件走 C2PA 标准路线。该机制覆盖 Anthropic 自家全部产品表面 —— Claude Platform API、Claude、Claude Code、Claude Cowork、Claude Tag —— 也覆盖 AWS、Google Cloud、Microsoft Foundry 等云厂商接入路径。适用范围不限于欧盟,Anthropic 表示该标记将跟随「Claude 在全球任何提供区域」生效。\n\n## 水印是怎么工作的\n\n文本侧:Claude 在生成阶段直接把不可见信号织进文本本身 —— 用户正常阅读看不到,也不改变语义、质量、可读性。**因为信号是文本的一部分,复制粘贴后仍会跟着走,部分编辑后仍可残留**。水印在模型层注入,所以无论来自哪条 Claude 产品线,输出都会带这个标记。\n\n文件侧:对支持的文件类型(包括 .svg、.png、.jpg 等),Claude 会附加基于 **C2PA(Coalition for Content Provenance and Authenticity)** 开放标准的数字签名出处元数据。第三方可凭签名验证文件是否经过 Claude 处理、以及元数据是否被篡改。\n\n第三方检测:Anthropic 将配套发布技术文档,帮助用户与第三方检测 Claude 的水印与 C2PA 元数据,但**具体的水印嵌入算法细节目前尚未披露**。\n\n## 已知边界(这点很关键)\n\nAnthropic 自己把限制写得很坦白:\n\n- **检测到水印 ≠ Claude 是原作者**:Claude 可能只是被用来校对、翻译、摘要、改写人类原创文本,输出仍可能携带水印。\n- **没检测到水印 ≠ 内容是人工写的**:被 Claude 处理前的旧模型、重度改写、翻译、混入其它段落、过短的文本都可能让信号不可测;文件元数据可能在截图、格式转换、再保存后丢失。\n- **老模型有过渡期**:2026 年 8 月 2 日之前发布的模型不立即覆盖,Anthropic 表示会在 AI 法案过渡期内补齐。\n- 因此,这个信号是「Claude 曾经处理过这段文字\u002F文件的提示」,而非「这是 AI 原创」的硬证据。\n\n## 来源与法律背景\n\n该政策落地于 Anthropic 已签署的 **欧盟 AI 法案第 50(2)条《AI 生成内容透明度实践守则》(Code of Practice on Transparency of AI-Generated Content)**。Anthropic 同时作为「生成式 AI 模型」与「生成式 AI 系统」两类提供者参与承诺。Claude 模型在欧盟 8 月 2 日及之后上线的版本自发布即支持标记;之前的模型进入过渡期补齐计划。\n\n## 行业影响\n\n第三方 AI 文本检测厂商 —— 给学校、出版商、雇主做「这是不是 AI 写的」概率推断的那批公司 —— 的商业模式面临重写:\n\n- **从「推断」转向「查信号」**:今天大多数检测器靠统计文本风格去猜;Anthropic 的做法是在自家输出里直接埋机器可读信号,平台与第三方只要跑检测器就能拿到确定性更高的判定。\n- **但短期冲击有限**:水印会被重写、翻译、混合抹掉,老模型不受限,旧版 Claude 输出的内容依然「裸奔」。\n- **如果 OpenAI、Google、Meta 跟进**:AI 文本检测可能从概率游戏升级为「出处平台」生意 —— 检测厂商要么转型做多源出处聚合,要么被模型厂商的原生信号挤压。\n\n## 评论:这件事比技术更重要\n\nAnthropic 这次真正的看点不是水印本身,而是 **「模型厂商把可追溯性写进了产品里」** 这个动作。复制粘贴 —— 这个 AI 时代最普通的行为 —— 即将多出一层隐性的可追溯性。短期对个人写作影响不大;但对学术诚信、新闻署名、合同生成、政务文档这些「必须人工署名」的场景,这是基础设施级的变化。\n\n至于 OpenAI、Google、Meta 是否会同步跟进,把 AI 文本变成自带「AI 痕迹」的格式,2026 下半年会有答案。如果三家都跟,这件事就会从 Anthropic 的合规动作,变成整个 LLM 行业的默认设置。\n\n参考资料:Anthropic 官方支持文档《How Claude marks AI-generated content》(support.claude.com)、TechStartups 2026-08-10 报道、Interesting Engineering 2026-08-11 报道、奇客 Solidot 2026-08-11 转载。","https:\u002F\u002Fsupport.claude.com\u002Fen\u002Farticles\u002F16266773-how-claude-marks-ai-generated-content","1fa87d30-d9f3-4752-b3be-0373933b3aaf",[11,15,18,21,24],{"id":12,"name":13,"slug":13,"description":14,"color":14},"c33b1bbc-d6ce-4f61-9d5d-1a0704a6a09b","ai-policy",null,{"id":16,"name":17,"slug":17,"description":14,"color":14},"1fcfaaf2-67de-43d3-9e35-5784852fec60","ai-safety",{"id":19,"name":20,"slug":20,"description":14,"color":14},"23544f6a-eea1-4f05-aa8d-749ca862d5d2","anthropic",{"id":22,"name":23,"slug":23,"description":14,"color":14},"dca4d0ab-7994-43a7-839e-7756fc77344a","claude",{"id":25,"name":26,"slug":26,"description":14,"color":14},"01598627-1ea6-4b27-a5d8-874971571a71","llm",[28],{"id":29,"lang":30,"title":31,"summary":32,"content":33},"619fec36-34b6-4dd5-b120-78fef66643fd","en","Anthropic's invisible Claude watermark survives copy-paste","Anthropic confirmed on August 10 that starting August 2, 2026, Claude models launched in or after the EU will embed machine-readable, user-imperceptible watermarks in generated text, and attach C2PA-compliant signed provenance metadata to supported image\u002FSVG files. The mechanism applies globally, not only in Europe, and stems from Anthropic's signing of the EU AI Act Article 50(2) Code of Practice on Transparency of AI-Generated Content. The goal: let platforms and third parties still detect Claude involvement after the text has been copied and pasted elsewhere.","## What happened\n\nOn August 10, Anthropic confirmed via its official support documentation that starting **August 2, 2026**, every Claude model launched in the EU after that date ships with built-in machine-readable content marking from day one. Text gets an embedded watermark; supported files get C2PA provenance metadata. The mechanism covers Anthropic's full product surface — Claude Platform API, Claude, Claude Code, Claude Cowork, and Claude Tag — and also extends to AWS, Google Cloud, and Microsoft Foundry. It is not EU-only: Anthropic says the marking applies wherever Claude is offered worldwide.\n\n## How the watermark works\n\n**On the text side**, Claude weaves an imperceptible signal directly into the generated text at generation time. Users do not see it; semantics, quality, and readability are unchanged. **Because the signal is part of the text, it travels with the content when copied and pasted elsewhere, and may survive partial editing.** The watermark is injected at the model level, so any output from any supported Claude surface carries the mark.\n\n**On the file side**, supported formats (.svg, .png, .jpg) get signed provenance metadata based on the **C2PA (Coalition for Content Provenance and Authenticity)** open standard. Third parties can verify whether a file was processed by Claude and whether the metadata has been tampered with.\n\n**Detection tooling**: Anthropic will publish technical documentation to help users and third parties detect both the text watermark and the C2PA metadata. The **specific watermark embedding algorithm has not been disclosed** yet.\n\n## Known limits (Anthropic is explicit about these)\n\n- **Detecting the watermark ≠ Claude authored it.** Claude may have been used only to proofread, translate, summarize, or rewrite human-written text. The output can still carry the mark.\n- **Failing to detect a watermark ≠ human authorship.** Old models before the cutoff, heavy edits, paraphrasing, translation, mixed passages, or very short snippets can erase the signal. File metadata can be stripped by screenshots, format conversions, or re-saving.\n- **Older models enter a transition period.** Models released before August 2, 2026 are not immediately covered; Anthropic says it is working on backfilling marking support during the EU AI Act transition window.\n- The signal is therefore a hint that **Claude processed the text or file**, not hard proof of AI authorship.\n\n## Source and legal basis\n\nThe policy implements the **EU AI Act Article 50(2) Code of Practice on Transparency of AI-Generated Content**, which Anthropic has signed as both a provider of generative AI models and a provider of generative AI systems. Claude models launched in the EU on or after August 2, 2026 are covered from launch; earlier models enter a transition plan.\n\n## Industry impact\n\nThird-party AI text detection vendors — the companies selling \"is this AI-written?\" probability inference to schools, publishers, and employers — face a rewrite of their business:\n\n- **From inference to signal lookup.** Today most detectors infer authorship from text statistics; Anthropic's approach embeds a machine-readable signal at the source, so platforms and third parties get a higher-confidence verdict from a deterministic check.\n- **But short-term disruption is limited.** Heavy edits, translations, mixing, and pre-cutoff-model output all weaken or remove the signal. Old-model output stays unmarked.\n- **If OpenAI, Google, and Meta follow suit**, AI text detection could shift from probabilistic guesswork to a \"provenance platform\" business. Detection vendors either pivot to multi-source provenance aggregation or get squeezed by the model makers' native signals.\n\n## Why this matters more than the tech itself\n\nThe real story is not the watermark algorithm; it is the fact that **a model vendor has now written traceability into the product itself**. Copy-paste — the most mundane behavior of the AI era — is about to gain a layer of implicit traceability. The personal-writing impact is small. But for academic integrity, news attribution, contract generation, and government documents — the scenarios where \"human authorship\" is a hard requirement — this is infrastructure-grade change.\n\nWhether OpenAI, Google, and Meta follow suit, and turn AI text into a format with built-in AI provenance by default, will be answered in late 2026. If all three follow, this moves from an Anthropic compliance move to an industry-wide default.\n\nReferences: Anthropic support doc *How Claude marks AI-generated content* (support.claude.com), TechStartups 2026-08-10, Interesting Engineering 2026-08-11, Solidot 2026-08-11.","anthropic-claude-invisible-watermark-eu-ai-act","2026-08-12T02:00:00Z","2026-08-11T16:04:57.493321Z","2026-08-11T16:04:57.493336Z",true,"agent",204,{"items":42},[43,48,53,58,63,68],{"id":44,"title":45,"news_slug":46,"published_at":47},"97c97b9c-e6e4-4982-aa57-0c0da814fb19","Anthropic 的欧盟答卷四小时即被撕开：Claude 文本水印为什么怕改写","claude-synthid-70-percent-threshold-bypass","2026-08-21T08:00:00+00:00",{"id":49,"title":50,"news_slug":51,"published_at":52},"ca53004e-9180-4b9d-b9db-337f2d20994b","Anthropic 给 Claude 文本加水印:欧盟 AI Act 第 50 条第一次有了「出厂级」答案","anthropic-claude-text-watermark-eu-ai-act","2026-08-11T21:48:00+00:00",{"id":54,"title":55,"news_slug":56,"published_at":57},"470b8663-3916-4bc5-ac3c-c592487c2873","Claude水印官宣4小时被破:开源去除工具走红,水印军备竞赛开场","claude-watermark-removal-tool","2026-08-20T19:30:00+00:00",{"id":59,"title":60,"news_slug":61,"published_at":62},"6b07def3-2b1c-4fe9-9922-0dd0038c149c","Anthropic 风险报告更新:Threat Model 1 升至「低」,Threat Model 2 维持「低」但信心下降","anthropic-risk-report-august-2026-update","2026-08-19T03:00:00+00:00",{"id":64,"title":65,"news_slug":66,"published_at":67},"a7f4cfad-874e-42b0-a84b-bd0ec57e8fdc","Anthropic 给 Claude 文本上不可见水印,接 SynthID-Text 走全球合规","anthropic-claude-invisible-text-watermark","2026-08-18T03:30:00+00:00",{"id":69,"title":70,"news_slug":71,"published_at":72},"a124851a-081e-44b6-9f20-f775404279e1","Claude 全球文本水印:Anthropic 把欧盟 AI Act 第 50 条做成\"全球默认\"","claude-text-watermark-eu-ai-act-global","2026-08-14T03:00:00+00:00"]