| Mobile Web

Search results for SynthID

AI & Enterprise

AI-generated content watermarks could undermine LLM safeguards

AI platforms are rolling out methods to insert watermarks into AI-generated content to comply with the European Union’s legal framework, but concerns are growing that the approach could make large language models more vulnerable to attacks. Lasso Security research found that SynthID-Text can affect not only word choice but also tool use and compliance with safeguards. In adversarial prompts, models sometimes followed instructions they would normally refuse.

AI & Enterprise

Anthropic discloses how Claude watermark works, says it disappears if fully rewritten

Anthropic has disclosed how an “invisible” digital watermark embedded in text generated by its AI chatbot Claude works, Japan’s ITmedia reported on Aug. 16. The method is based on Google DeepMind’s SynthID-Text and leaves statistical patterns through token selection without adding extra characters. Anthropic plans a global rollout tied to an EU AI Act transparency code of practice. It said light edits do not fully remove the mark, but full rewrites do.

AI & Enterprise

Anthropic clarifies after Claude watermark controversy over AI labeling

Anthropic has begun inserting hard-to-detect watermarks into outputs from some Claude models, raising concerns over copyright and privacy. The watermark can remain after copy-and-paste and may persist after edits, and Anthropic says it is meant to distinguish human and AI writing. Critics warn the company could act as sole arbiter, while developers question impacts on text quality and on lightly edited human writing. Anthropic says it will provide a free API and that the mark indicates processing, not authorship.