Chinese language regulators possible discovered from the EU AI Act, says Jeffrey Ding, an assistant professor of Political Science at George Washington College. “Chinese language policymakers and students have mentioned that they’ve drawn on the EU’s Acts as inspiration for issues previously.”
However on the identical time, a few of the measures taken by the Chinese language regulators aren’t actually replicable in different nations. For instance, the Chinese language authorities is asking social platforms to display the user-uploaded content material for AI. “That appears one thing that could be very new and may be distinctive to the China context,” Ding says. “This may by no means exist within the US context, as a result of the US is known for saying that the platform just isn’t liable for content material.”
However What About Freedom of Expression On-line?
The draft regulation on AI content material labeling is looking for public suggestions till October 14, and it might take one other a number of months for it to be modified and handed. However there’s little cause for Chinese language firms to delay getting ready for when it goes into impact.
Sima Huapeng, founder and CEO of the Chinese language AIGC firm Silicon Intelligence, which makes use of deepfake applied sciences to generate AI brokers, influencers, and replicate dwelling and lifeless folks, says his product now permits customers to voluntarily select whether or not to mark the generated product as AI. But when the regulation passes, he may need to vary it to obligatory.
“If a characteristic is elective, then almost definitely firms gained’t add it to their merchandise. But when it turns into obligatory by regulation, then everybody has to implement it,” Sima says. It is not technically tough so as to add watermarks or metadata labels, however it would improve the working prices for compliant firms.
Insurance policies like this could steer AI away from getting used for scamming or privateness invasion, he says, but it surely might additionally set off the expansion of an AI service black market the place firms attempt to dodge authorized compliance and save on prices.
There’s additionally a advantageous line between holding AI content material producers accountable and policing particular person speech by way of extra refined tracing.
“The massive underlying human rights problem is to ensure that these approaches do not additional compromise privateness or free expression,” says Gregory. Whereas the implicit labels and watermarks can be utilized to determine sources of misinformation and inappropriate content material, the identical instruments can allow the platforms and authorities to have stronger management over what customers put up on the web. In truth, issues about how AI instruments can go rogue has been one of many essential drivers of China’s proactive AI laws efforts.
On the identical time, the Chinese language AI trade is pushing again on the federal government to have more room to experiment and develop since they’re already behind their Western friends. An earlier Chinese language generative-AI regulation was watered down significantly between the primary public draft and the ultimate invoice, eradicating necessities on id verification and decreasing penalties imposed on firms.
“What we have seen is the Chinese language authorities actually attempting to stroll this advantageous tightrope between ‘ensuring we keep content material management’ but in addition ‘letting these AI labs in a strategic house have the liberty to innovate,’” says Ding. “That is one other try to try this.”
