TikTok has pulled its controversial AI remix setting, the company confirmed to CNET on Thursday. This is a stark change of course for a company that's been on a long journey to integrate AI into every part of its platform.

TikTok had introduced the setting to test an AI-powered meme maker inspired by videos and clips on the platform, the company told CNET earlier this week. The new AI feature, called the meme remixer, would allow anyone who sees your TikTok video to create an AI-generated image from it, changing the face or background. The hidden setting automatically granted users permission to remix your posted content with AI, unless you explicitly toggled it off. 

Creators who spotted the setting on the social media app complained about having their videos automatically opted in and worried about how their data would be used. 

For some TikTokers, it felt like an invasion of privacy and another attempt by social media platforms to stealthily alter original content. 

"It shouldn't be that hard to allow us to opt out in one toggle setting," creator Sean Szolek-Van Valkenburgh said in his video. While creators knowingly hand over some rights to their content when posting to platforms, there should be an option to limit how much of that is exploited by AI, he said.

TikTok has been introducing AI features to its platform for several years now. You might've noticed the new translucent icon above a profile picture when you're swiping through videos -- that's Tako (pronounced "taco"), TikTok's AI assistant. 

Like other social media platforms, TikTok has faced backlash for integrating AI, with creators panicking that AI slop is drowning out their original content, along with environmental, legal and ethical concerns. But that hasn't stopped companies like TikTok, Snapchat and Meta, which own Instagram and Facebook, from charging into AI.

Remixable content 

Before TikTok pulled it, the AI remix setting was on by default. There was a toggle you could switch off to prohibit others from creating AI memes from your clips. The setting worked similarly to those that let you control whether people can duet or stitch your videos. 

So, if the setting wasn't switched off, and you posted a video of yourself at a new coffee shop, a commenter could hypothetically use the meme remixer to type in a custom prompt and create an AI image featuring your likeness on a beach or in another scenario. These images would then get shared in the comments of the original video. 

Though you could turn off AI remixing on individual videos, there was no way to opt out at an account level. TikTok users had to turn this setting off for every video. The experimental meme tool has also been paused as TikTok evaluates feedback.

TikTok had said that if you allowed your videos to be remixed with AI, they wouldn't be used to train its AI models. TikTok said the toggle had no impact on how TikTok used or interacted with your content. 

But given the black box nature of how tech companies develop AI, there's no way for individual creators to confirm this. 

AI content on TikTok 

In the company's general AI policy, it says any content edited or entirely created by AI has to adhere to its community guidelines. 

TikTok's rules around edited media specifically outlaw content that "misleads" users into believing something fake is real, particularly when it depicts "fake authoritative sources or crisis events, or falsely shows public figures in certain contexts." This suggests that deepfakes would be prohibited. Child sexual abuse material is also explicitly banned. 

AI-edited content shared on TikTok has an invisible watermark, compliant with the C2PA standard. But with newer, more powerful AI image and video models, it's hard for folks to tell whether something is real or AI-generated.

While policies can look good on paper, it's their enforcement that matters. One creator, Georgie, who goes by the handle soupytime, called out that it's already easy for people to repost or steal viral videos, and that the new AI tool would make it even easier to create deepfakes of well-known creators without their consent. 

This can be dangerous, as we've seen with Grok AI on X. Without proper precautions, it's easy for social media users to take photos and videos that people share online and turn them into abusive, illegal content.