Anthropic has embedded invisible cryptographic fingerprints in every text response from the Claude API since August 2, 2026, steering token choices with a secret key that only Anthropic and approved p…
Lasso Security found that watermarking large language models changes AI agent behavior, with six of seven tested models scoring lower on tool-call tasks and four of those declines statistically signif…
Research reported by Ars Technica found that Google's SynthID-Text watermarking can alter LLM responses to adversarial prompts and increase the risk of safety guardrail bypasses, affecting tool invoca…
Anthropic announced that future Claude models will embed an invisible watermark based on Google DeepMind's SynthID-Text, a deployment with regulatory relevance under Article 50(2) of the EU AI Act, wh…
Anthropic will integrate Google DeepMind's SynthID-Text watermarking into every new Claude model starting August 2, 2026, with existing Claude models retrofitted by December 2, 2026, the company said.…
New research from Lasso Security found that Google's SynthID-Text watermarking can alter not only word selection but also the tools a large language model invokes and whether it adheres to or disregar…
A paper submitted to arXiv on 9 Sep 2026 argues that the AI text watermarks now required under Article 50 of the EU AI Act, which took effect on August 2, 2026, cannot currently be verified by users o…
Anthropic has introduced an invisible text watermark for Claude-generated text and digitally signed provenance metadata for generated files, aiming to increase transparency in AI content. The watermar…
A data analysis by Louis Abraham shows that Claude's GitHub pull request descriptions have become increasingly repetitive, with one of eight writing styles growing from 1.0% of the corpus in early 202…
Anthropic has signed the European Union's Code of Practice on Transparency of AI-Generated Content and is implementing machine-readable watermarking across all output from its Claude models worldwide,…
A developer has released a minimal implementation of LLM watermarking, inspired by Anthropic's upcoming 'is it AI' API and Google DeepMind's SynthID-Text. The code, written in Python using PyTorch and…
Starting August 2, 2026, every Claude model Anthropic ships embeds an invisible statistical watermark into every text response, including API calls, with no opt-out. The watermark, based on Google Dee…
A quiz by software engineer Sean Godecke found that readers could not identify AI watermarked text, with 73 participants averaging 3.4 out of 10 correct guesses, close to random chance. The quiz used …
A new interactive quiz challenges users to distinguish between watermarked and non-watermarked AI-generated text, using Qwen2.5-32B-Instruct and SynthID-Text, the watermarking method behind Gemini and…
Google DeepMind's SynthID-Text watermarking, described in an October 2024 Nature paper, embeds a statistical bias in token selection to mark AI-generated text, and as of August 2, 2026, Claude and Gem…
Anthropic revealed on August 14 that its new text watermarking for Claude is based on Google DeepMind's SynthID-Text and applies to text longer than 200 tokens, with no practical impact on output qual…
Google DeepMind researchers introduced SynthID-Text, a production-ready text watermarking scheme for large language models that preserves text quality and enables high detection accuracy with minimal …
Anthropic has begun embedding a statistical watermark into text generated by supported Claude models, a move designed to comply with the European Union's AI Act Article 50, which took effect on August…
Anthropic has begun using Google DeepMind's SynthID-Text to invisibly watermark text generated by its Claude AI models, aiming to comply with the EU AI Act's transparency mandates. The watermarking wo…
Anthropic has begun watermarking all text outputs from new Claude models globally with an invisible statistical watermark, meeting the EU AI Act's August 2 deadline, with no opt-out and no regional ge…