Tech News
← Home  ·  All topics

Ai Watermarking

5 GoKawiil briefs on this topic · RSS for ai watermarking

Anthropic's text watermarking sparks quality debate after EU AI Act rules take effect

Following the EU AI Act's transparency requirements, major AI labs including Anthropic have begun watermarking AI-generated text to signal its origin. This has triggered a dispute among commentators, including writer John Gruber and Daniel Jalkut, over whether embedding these hidden markers degrades the clarity or quality of the generated text, with Anthropic itself maintaining that watermarking does not affect output quality, citing Google's SynthID-Text research.

Nature correspondence warns AI text watermarks are vulnerable to cheap attacks

Responding to a Nature news story on Anthropic's invisible watermarking of AI-generated text, researchers point out that such watermarks can be undermined. They cite a 2024 study in which attackers inferred a watermark's pattern by querying the model, then spoofed or removed it for under $50, succeeding roughly 80% of the time.

Google launches standalone SynthID.com site to detect AI watermarks from multiple vendors

Google has opened a public website, SynthID.com, that lets anyone check images, video, or audio for invisible SynthID watermarks. The tool now recognizes watermarks from partner platforms including OpenAI, Nvidia, and Kakao, with Apple support coming soon, rather than only detecting Google's own Gemini-generated content as before. Google says Gemini alone has embedded SynthID in over 180 billion images and videos and 240,000 years of audio.

Google opens SynthID Detector website to all users

Google has made SynthID Detector, previously limited to journalists and media professionals, publicly available at synthid.com. The free tool detects AI watermarks in images, video and audio from Google, OpenAI, NVIDIA and Kakao, requiring users to log in with a Google, OpenAI or Apple account. Support for detecting Apple Intelligence-created images is expected soon.

Google DeepMind unveils SynthIDBio watermark for AI-designed proteins

Google DeepMind researchers in London have built a watermarking system, described in Nature, that embeds a hidden statistical signal into both the amino-acid sequence and 3D structure of AI-designed proteins. The tag, called SynthIDBio, is meant to mark proteins as machine-generated without impairing their biological function, allowing them to be identified later in databases.