ICYMI 👀: Anthropic will start watermarking AI-generated text via upcoming models released in the EU after the 2nd of August.
The same capability for existing models is in the works as well.
"Claude uses two complementary techniques to mark content generated and processed by Claude: watermarks embedded in text, and signed provenance metadata attached to files."
Does it mean that EU users will start experiencing model release delays due to this?
Google has had the same tech in place for quite a while. OpenAI developed a similar solution some time back; however, it was not applied to their models.
Can we instead agree to mark human-generated content? That would be much more efficient in the long run.
Anthropic says new Claude models will embed invisible watermarks in all generated text, everywhere Claude is offered. The watermark is part of the text, it isn'...