待翻译：Anthropic pledges to embed watermarks to help discern AI slop in sop to EU

AI 服务暂时不可用，以下为来源正文，待恢复后补全翻译。

Anthropic pledges to embed watermarks to help discern AI slop in sop to EU EU rules cited as reason for effort to trace AI output ancestry Thomas Claburn Thomas Claburn AI AND SOFTWARE REPORTER Published tue 11 Aug 2026 // 01:27 UTC Anthropic will embed watermarks in the text and files generated by future models it launches in the EU, as part of its effort to comply with content and transparency rules in the bloc's AI Act. "Generated text will carry embedded watermarks, and generated files will include digitally signed provenance metadata where supported," the company said in a help document published on Monday. The AI biz is also working on models it has already released, to add output marking during the transition period allowed under EU law. REG AD "Marking will apply to output from supported models wherever Claude is offered, worldwide," the company said. REG AD The move may further amplify the appeal of open weight models and alienate Claude customers, who don't necessarily want consumers of their AI-generated content to know its provenance. In the past, Claude users have expressed frustration with pricing changes, reliability issues, and model safeguards that have hindered legitimate work in the name of safety. Claude users appear to be skeptical that a text-based watermarking scheme will work. Researchers have already demonstrated that image-based watermarking can be undone. Anthropic intends to apply marks to output from covered Claude models on the Claude Platform (API), Claude, Claude Code, Claude Cowork, and Claude Tag. This also includes third-party providers of Anthropic models like AWS, Google Cloud, and Microsoft Foundry. The Ai biz expects to provide details about how people can detect Claude's marks, as required under EU law, in forthcoming documentation. "When a supported Claude model generates text, it weaves an imperceptible watermark directly into the text itself," the biz said. "You won’t see it, and it doesn’t change the meaning, quality, or readability of Claude’s response. MORE CONTEXT Hey, big spender – OpenAI has a new SKU just for you DEF CON hackers add new muscle to water utility protection The future is for billionaires – the rest of us will get open weight AI models, maybe Zuck rekindles open weights Llama drama with Muse Glimmer "Because the watermark is part of the text, it will travel with the text when it’s copied and pasted elsewhere, and may persist through some editing. Watermarking will be applied at the model level, which means it will be present no matter which Claude product or surface the text comes from." Absent examples or technical documentation, it's unclear how Anthropic will make the watermarks hard to remove. But it should not be difficult to create an optical character recognition system that strips or omits obscure marks from Claude-generated text. Anthropic's insistence that its text marking scheme "doesn't change the meaning" of Claude's output appears to preclude using word choice and placement as a text provenance identifier – a technique Apple has reportedly used to catch those who would violate its employee secrecy agreements. REG AD As for marks attached to Claude-created files, Anthropic says it will rely on signed provenance metadata that conforms to the C2PA standard, for which there are already open source removal tools. Anthropic itself is already hedging about the utility of its marking method, noting that detected marks are not conclusive evidence that Claude produced the content and that the absence of marks cannot guarantee that AI wasn't involved in the creation of a particular piece of content. But perhaps the scheme is good enough to count as legal compliance. ®