According to CoinMeta, Anthropic has announced that it will add machine-readable invisible watermarks to the text generated by its AI model Claude. This change is aimed at fulfilling the transparency requirements of the EU's AI legislation. The watermarks will be embedded directly when the model generates text and are invisible to the naked eye; they will also be retained upon copying and pasting. These watermarks utilize Google DeepMind's SynthID technology and do not involve the use of zero-width characters or hidden code. The addition of watermarks will not increase the token count and will have almost no impact on generation speed or price. Although some users have canceled their Claude subscriptions as a result, the company indicates that there has been no overall increase in unsubscriptions. Anthropic also plans to make a watermark detection API available for users and third parties to check the text themselves.