Watermarking for Language Models

The EU AI Act has certain implications where companies will have to disclose their data sets, et cetera. But I think this watermarking technique is really interesting because all of a sudden, if you could just watermark your data and then look at the outputs at the model, you could essentially have a way for distribution channels to look for that watermark as well. So particularly excited about where the watermarking efforts are going internally. And we've gotten quite a few large enterprises really interested in this as well,. Would it also be possible to do something where, and if anybody can hear my dog, Ringo barking in the background, I apologize.