Anthropic 水印在旧版 Claude 上易被绕过

Nano Banana Pro,由 THE DECODER 提供提示词
更新,2026 年 11 月 4 日:
Anthropic 再次向 API 客户发送邮件,说明第二轮水印回溯加装。自 2026 年 11 月 16 日起,Claude Opus 4.6、Claude Opus 4.5、Claude Sonnet 4.5 和 Claude Haiku 4.5 都将加入文本水印。在 Amazon Bedrock 上,推送可能要多花几天。
Claude Opus 4.6 原本属于第一批,本应在 11 月 3 日获得水印。Anthropic 明确确认推迟到 11 月 16 日,但没有给出原因。第一批的另外三个模型——Claude Mythos 5、Claude Opus 4.7 和 Claude Sonnet 4.6——没有出现在这封新邮件中,因此它们的推送很可能已按计划在 11 月 3 日进行。
Claude Opus 4.5、Claude Sonnet 4.5 和 Claude Haiku 4.5 是三个未出现在此前公告中的模型。这一批的四个模型都有同样的采样限制:只有当 temperature、top_p 和 top_k 设为默认值时,水印才会生效。在第一批中,这一限制只涉及四个模型中的两个(Opus 4.6 和 Sonnet 4.6)。
原文,2026 年 10 月 31 日:
Anthropic 正将其文本水印扩展到另外四个 Claude 模型,但 API 用户只要调整一个采样参数,就能在其中两个模型上避开这一标记。
在发给 API 客户的邮件中,Anthropic 表示将为 Claude Mythos 5、Claude Opus 4.7、Claude Opus 4.6 和 Claude Sonnet 4.6 添加水印。像 Claude Opus 5.5 这样较新的模型已经带有该标记,而在 Amazon Bedrock 上的推送可能要多花几天。
Anthropic 将这种水印称为"用词中不易察觉的模式",不会改变含义、质量或可读性。该公司表示,它不会增加额外的字符或 Token,不影响定价或请求与响应的格式,对输出速度的影响也微乎其微。Anthropic 还表示,该水印"不包含任何关于用户、用户所属组织或其与 Claude 对话的信息"。
Anthropic 仅在两款较旧模型的默认设置下应用该水印
Anthropic 表示,在 Claude Opus 4.6 和 Claude Sonnet 4.6 上,只有当开发者将三个采样参数保持在默认值时,水印才会生效。这三个参数 temperature、top_p 和 top_k 用于控制模型在选词时使用的随机程度。只要对其中任何一个做细微调整,就足以避开该水印。Anthropic 在邮件中警告称:"如果你依赖该水印来满足自身的合规要求,请将这些设置保持在默认值。"
该邮件并未对 Claude Mythos 5 或 Claude Opus 4.7 提及这一限制条件,这暗示该水印在这两款模型上无论采样设置如何都会生效。但较旧模型上的这一漏洞仍然值得注意,因为 Opus 4.6 仍是 Anthropic 最强的写作模型,甚至领先于 Opus 5.5。
Anthropic 有一个处于私密预览阶段的检测 API,用于验证该水印,并在其内容标注信息页面上提供了更多细节。
Anthropic 为较旧的 Claude 模型补加水印,以赶上欧盟《AI 法案》的截止期限
Anthropic 将欧盟《人工智能法案》12 月的截止期限作为其现在为较旧模型添加水印的原因。2026 年 8 月 2 日之后发布的每一款模型,在发布首日就带上了水印;而较旧的模型则分阶段获得该功能。由于水印嵌入在模型层面,它在托管这些模型的所有平台上都有效,包括 Amazon Web Services、Google Cloud 和 Microsoft Foundry。客户无法将其关闭。
Anthropic 采用了 Google DeepMind 的 SynthID 文本方法的变体,该方法通过调整选词背后的随机性,产生一种在统计上可检测的模式。OpenAI 于 2026 年 10 月推出了自己的水印系统 textGrain,但它仅对欧盟地区的 ChatGPT 和 Codex 用户默认启用。全球的 API 客户可以自行开启或关闭,且默认处于关闭状态。Google DeepMind 已将用于文本的 SynthID 开源,并将其内置到 Gemini 系列模型中。
水印是否真的完全不影响文本质量,仍存在争议。博主 John Gruber 认为,没有哪两个同义词的含义完全相同,因此依据水印密钥而非语义契合度来选词,必然会导致精确性下降。法律行业媒体 Artificial Lawyer 则针对律所提出了另一项担忧:已禁止使用 AI 的客户可能会发现水印痕迹,而同一份文档中叠加的多个水印也可能带来麻烦。
没有炒作的 AI 新闻——由人工精选
译文由英文原文机器翻译,术语按行业惯例统一,可在文末展开原文核对。
查看英文原文(Anthropic's text watermark is easy to bypass on several older Claude models)
Nano Banana Pro prompted by THE DECODER
Update, November 4, 2026:
Anthropic emailed API customers again about a second round of watermark retrofits. Starting November 16, 2026, Claude Opus 4.6, Claude Opus 4.5, Claude Sonnet 4.5, and Claude Haiku 4.5 will all get text watermarking. On Amazon Bedrock, the rollout may take a few extra days.
Claude Opus 4.6 was originally part of the first wave and should have gotten the watermark on November 3. Anthropic explicitly confirms the delay to November 16 but gives no reason. The other three first-wave models, Claude Mythos 5, Claude Opus 4.7, and Claude Sonnet 4.6, don't show up in the new email so their rollout likely went ahead on November 3 as planned.
Claude Opus 4.5, Claude Sonnet 4.5, and Claude Haiku 4.5 are three models that weren't part of the previous announcement. All four models in this wave share the same sampling restriction: the watermark only applies when temperature, top_p, and top_k are set to their defaults. In the first wave, that limit only hit two of four models (Opus 4.6 and Sonnet 4.6).
Original article, October 31, 2026:
Anthropic is expanding its text watermark to four more Claude models, but API users can dodge the mark on two of them by tweaking a single sampling parameter.
In an email to API customers, Anthropic said it will watermark Claude Mythos 5, Claude Opus 4.7, Claude Opus 4.6, and Claude Sonnet 4.6. Newer models like Claude Opus 5.5 already carry the mark, and the rollout on Amazon Bedrock may take a few extra days.
Anthropic calls the watermark an "imperceptible pattern in word choice" that doesn't change meaning, quality, or readability. The company says it adds no extra characters or tokens, doesn't affect pricing or request and response formats, and barely slows output speed. Anthropic also says the watermark "contains no information about the user, their organization, or their conversations with Claude."
Anthropic applies the watermark only at default settings on two older models
On Claude Opus 4.6 and Claude Sonnet 4.6, Anthropic says the watermark kicks in only when developers leave the three sampling parameters at their default values. Those parameters, temperature, top_p, and top_k, control how much randomness the model uses when picking words. A minor tweak to any one of them is enough to dodge the watermark. Anthropic warns in its email that "If you rely on the watermark for your own compliance, keep these settings at their defaults."
The email doesn't mention that caveat for Claude Mythos 5 or Claude Opus 4.7, which suggests the watermark applies to those models regardless of sampling settings. But the loophole on the older models is still notable, since Opus 4.6 remains Anthropic's best writing model, ahead of even Opus 5.5.
Anthropic has a detection API in private preview for verifying the watermark and offers more details on its content labeling information page.
Anthropic retrofits older Claude models to meet the EU AI Act deadline
Anthropic points to the EU AI Act's December deadline as the reason it's watermarking older models now. Every model released after Aug. 2, 2026, shipped with the watermark on day one, while older models are getting it in stages. Since the watermark is embedded at the model level, it works on every platform that hosts these models, including Amazon Web Services, Google Cloud, and Microsoft Foundry. Customers can't turn it off.
Anthropic uses a variant of Google DeepMind's SynthID text method, which tweaks the randomness behind word selection to produce a statistically detectable pattern. OpenAI launched its own watermarking system textGrain in October 2026, but it only activates by default for ChatGPT and Codex users in the EU. API customers worldwide can flip it on or off, and it ships disabled. Google Deepmind open-sourced SynthID for text and built it into the Gemini models.
Whether watermarking truly leaves text quality untouched is disputed. Blogger John Gruber argues that no two synonyms mean exactly the same thing, so picking words by a watermark key instead of semantic fit inevitably degrades precision. The legal trade outlet Artificial Lawyer raises a separate concern for law firms, where clients who've banned AI use could discover the mark, and multiple watermarks layered inside one document could create complications.
AI News Without the Hype – Curated by Humans
原文为英文,正文由自动抓取并保留原文。
原文发布于 The Decoder(Anthropic's text watermark is easy to bypass on several older Claude models),标题、摘要与正文由机器翻译,术语按行业惯例统一,未经人工逐句校订。