POLICY18 AUG 2026 · 09:15 UTC · 2 MIN READ

How do AI text watermarks work?

Anthropic's Claude uses a version of the SynthID-Text approach to apply invisible watermarks to generated text, complying with Europe's AI transparency rules. The watermarking technology creates detectable patterns using wording probabilities, leaving a pattern in the text that is undetectable to readers but detectable to those with a key.

POLICY DESK
How do AI text watermarks work?
Anthropic's Claude uses a version of the SynthID-Text approach to apply invisible watermarks to generated text, complying with Eur…
How do AI text watermarks work?

So, Anthropic's finally shed some light on how it's planning to use invisible watermarks on text generated by Claude, all in an effort to comply with those stringent AI transparency rules in Europe. It turns out that Claude's text marking system is actually based on the SynthID-Text approach - an open-source watermarking tech developed by Google DeepMind. This clever technology creates detectable patterns using wording probabilities, which allows the generated text to be identified as artificially generated or manipulated.

The whole point of introducing this watermarking feature is to meet Anthropic's obligations under the European Union's AI Act, which demands that synthetic audio, image, video, and text include machine-readable marks that enable the content to be detected as artificially generated or manipulated. Thankfully, Anthropic assures us that these text watermarks won't make Claude more expensive for users, nor will they have any practical impact on the quality or content of Claude's outputs - that's a relief!

Now, here's how the watermarking process actually works: it uses low-stakes choices in generated text to leave a pattern in Claude's responses. For instance, when generating a sentence, the model might choose between two words that are equally likely to follow the previous words - in cases like this, the choice is settled by a random number. The watermarking tech uses these low-stakes choices to leave a pattern in the text, which is undetectable to the reader but detectable to anyone who has a key that encodes it.

As Anthropic points out, the EU's AI transparency requirements are going to impact other major AI developers too, so Claude won't be the only model introducing text watermarks. Other big players like Google and OpenAI will also be subject to the law's requirements, and may well introduce similar watermarking technologies to their AI models. So, users of AI models can expect to see more widespread adoption of text watermarks in the future, as companies scramble to comply with the EU's AI transparency rules.

Source: The Verge

OpenAI Begins Embedded Assessments With Four Independent Labs
policy

OpenAI Begins Embedded Assessments With Four Independent Labs

OpenAI has launched its first embedded assessments, granting four independent labs-Apollo Research, METR, Transluce, and Goodfire-deep, sustained access to model checkpoints, source code, and training data. Unlike standard audits, evaluators work on-site for months. Each lab focuses on distinct risks: covert behavior, interpretability during training, reinforcement learning misalignment, and post-training evaluation quality. Overseen by OpenAI’s Safety and Security Committee, the progr

Google Launches Gemini Decisions API for Multimodal Evaluation
models

Google Launches Gemini Decisions API for Multimodal Evaluation

Google launched the Gemini Decisions API on November 11, 2026, enabling developers to evaluate up to 1,000 multimodal questions-spanning text, images, audio, video, and PDFs-in a single batch call. The API, built on the Gemini Decision 1 model, supports yes/no, multiple-choice, and rubric-based scoring with calibrated confidence scores, eliminating the need for separate preprocessing pipelines for each media type. Pricing, benchmarks, and latency details remain undisclosed.

OpenAI Launches ChatGPT Wallet for Agent Payments
tools

OpenAI Launches ChatGPT Wallet for Agent Payments

OpenAI launched ChatGPT Wallet on November 13, 2026, enabling AI agents to complete purchases for ChatGPT Pro subscribers in the US. Built with American Express, Mastercard, and Visa, the system tokenizes card details so merchants never see real credentials. Users approve each transaction via biometrics after reviewing items, totals, and the selected card. OpenAI is rolling out access narrowly to refine compatibility with existing checkout and fraud-prevention systems.

Share this post
Report