Who's doing what with watermarking AI content?

Anthropic is adopting Google DeepMind's SynthID-style text watermarking for Claude, with existing models transitioning over the coming months and a text-detection API planned, while Google has open-sourced SynthID Text. The EU AI Act's transparency rules, effective 2 August 2026, are driving adoption, with Anthropic, Google, OpenAI, Meta, Microsoft, Mistral, and Cohere signing the EU code of practice, though OpenAI, Meta, Microsoft, Mistral, Cohere, and xAI have not yet announced text watermark implementations.

Who’s doing what with watermarking AI content? /blog/2026/8/17/whos-doing-what-with-watermarking-ai-content Anthropic is bringing SynthID-style text watermarking to Claude https://www.anthropic.com/news/claude-text-watermark , John Gruber has some objections https://daringfireball.net/2026/08/anthropics watermark text adulteration in claude is a perversion of writing , and James Padolsey has the clearest explanation I’ve seen of how it works https://declaude.org/watermarking/ . This whole discussion got me wondering what everyone else is doing. Here is where the major AI companies currently stand: Google DeepMind developed SynthID, initially for generated images. In 2024 it extended SynthID https://deepmind.google/blog/watermarking-ai-generated-text-and-video-with-synthid/ to text in the Gemini app and web experience, then open-sourced SynthID Text https://developers.googleblog.com/en/evolving-the-responsible-generative-ai-toolkit-with-new-tools-for-every-llm/ . That is why Anthropic can adopt a version for Claude, and so can anyone else. Google’s public Gemini and SynthID verification tools https://deepmind.google/models/synthid/ currently accept images, audio and video, but not pasted text. Google of course has another similarly named product called the AI Content Detection API https://docs.cloud.google.com/gemini-enterprise-agent-platform/models/ai-content-detection . That one examines images from Google and third-party models using pixel-level artifacts and other statistical clues. Despite the broad name, its current documentation only lists JPEG, PNG and WebP inputs. Anthropic Future Claude https://www.anthropic.com/news/claude-text-watermark models will use a version of Google’s SynthID Text. Existing models will move over during the coming months. C2PA metadata is also included with some supported files. Anthropic says a text-detection API is coming soon https://www.anthropic.com/news/claude-text-watermark . OpenAI OpenAI has not released a text watermark, although it says it intends to expand provenance support to text. Supported images use C2PA and SynthID, while supported audio uses SynthID. OpenAI Verify https://help.openai.com/en/articles/8912793-provenance-signals-content-credentials-synthid-in-openai-generated-content currently checks supported images and audio. Meta I couldn’t find an announced text watermark for Meta AI or Llama. Meta’s new Content Seal https://ai.meta.com/blog/introducing-muse-image-muse-video-msl/ marks generated images, with video support planned. Microsoft Microsoft has not announced a text implementation. Microsoft 365 can add watermarks and provenance metadata to generated images, video and audio https://support.microsoft.com/en-US/Privacy/watermarks-on-content-generated-or-altered-by-using-ai-in-microsoft-365-at-work-or-school , but there is no general text checker. Mistral and Cohere Both have committed to the EU Code https://digital-strategy.ec.europa.eu/en/policies/code-practice-ai-generated-content , but neither has publicly described a text-watermarking implementation that I could find. xAI xAI has not announced a text watermark. Grok adds watermarks to generated images and video, but offers nothing public for checking text. The sudden activity is being driven in part by the EU AI Act. Its transparency rules began applying on 2 August 2026 and require generated text, images, audio and video to be marked in a machine-readable form where technically feasible. Anthropic, Google, OpenAI, Meta, Microsoft, Mistral and Cohere have all signed the relevant section of the EU code of practice https://digital-strategy.ec.europa.eu/en/news/strong-backing-code-practice-transparency-ai-generated-content . What about open-weight models? Google has open-sourced SynthID Text https://github.com/google-deepmind/synthid-text , and it is available in Hugging Face Transformers. That means I could download an open-weight model such as Qwen3-8B https://huggingface.co/Qwen/Qwen3-8B , run it at home, choose my own keys and produce watermarked text. But you can just as easily run the same model without SynthID; open-weights and watermarking via this technique don't really jibe.