In response, the guidelines regulate the labeling of AI-generated online content throughout its production and dissemination processes, requiring providers to add visible marks to their content in appropriate locations.

My understanding is that this is meant more as a set of legal guidelines… I’m not a legal scholar, but since China has a history of enforcing certain information-related laws I’d assume they can “legally” enforce it

On the technical side… there is a subfield of LLM research that focuses on “watermarking” or ensuring that LLM-generated outputs can be clearly identified, so I guess in theory it might be enforceable

In practice as to whether it will actually be ensured… who knows (facepalm

This can probably only be enforced legally, not technologically…

By prosecuting people/companies who does not label their AI generated content?

You can use those AI image detectors, but they’re not foolproof and the companies that produce AI image generators will always be trying to improve their product so that it can’t be detected. It’s a place to start, though.

Realistically, it’ll be fear of prosecution. The real question is who they would prosecute? If a random individual posts an AI image on WeChat that isn’t tagged as AI, is the Chinese government going to fine them? Or will they go after WeChat for allowing unlabeled images to be shared?

They can shut down companies that make AI image generation software if the software doesn’t automatically tag the image as AI generated, but that will only apply to companies within China. They’ll probably look to ban access to all AI image generation that isn’t housed within China (if they don’t already).

And this will only ever apply to images shared publicly. If someone send a private message to someone else with an untagged AI image, nothing will happen.

You don’t. You just say you do so the more docile members of your population don’t notice when you start pushing untagged AI-generated propaganda