TikTok Introduces AI Content Controls and Invisible Watermarking to Combat AI Slop

TikTok now lets you choose how much AI-generated content you want to see | TechCrunch

TikTok, an app that was once just a place for user-generated content, is launching a new setting that lets users choose how much AI-generated content they want to see in their "For You" feed. The company is also introducing more advanced labeling technologies for AI-generated content. The new AI-generated content (AIGC) control is rolling out within the app's "Manage Topics" tool, which lets users choose what they see on TikTok. "Manage Topics already enables people to adjust how often they see content related to over 10 categories like Dance, Sports, and Food & Drinks," TikTok explained in a blog post. "Like those controls, the AIGC setting is intended to help people tailor the diverse range of content in their feed, rather than removing or replacing content in feeds entirely." The move comes as companies like OpenAI and Meta are embracing AI-only feeds. In September, Meta released Vibes, a new feed for sharing and creating short AI-generated videos. A few days after Meta's launch, OpenAI released Sora, a social media platform for creating and sharing AI-generated videos. Since Sora's launch, realistic AI-generated videos have been posted to TikTok. Additionally, many TikTok users are leveraging AI to create visuals for posts about other topics, like history or celebrities. TikTok says that with the new AI-generated content control, users who want to see less of this sort of content can now dial things down, while those who enjoy it can choose to see more of it. You can access the new capability by going into your Settings, selecting "Content Preferences" and then clicking the "Manage Topics" option. Then, you can move the slider for different topics, including AI-generated content, to adjust how much you do or don't want to see that sort of content in your For You feed. The change is rolling out in the coming weeks, TikTok says. To improve its ability to label AI-generated content, TikTok is now testing a technology called "invisible watermarking." TikTok already requires people to label realistic AI-generated content and uses a cross-industry technology called Content Credentials from C2PA, which embeds metadata into content that lets it and other platforms know when something is AI-generated. However, TikTok notes that these labels can be removed when content is reuploaded or edited on other platforms. With the new "invisible watermarks," TikTok will add another layer of safeguards by using a watermark that only it can read. That means it'll be harder for others to remove it. TikTok will start adding invisible watermarks to AI-generated content made with TikTok tools like AI Editor Pro. It's also adding them to content uploaded with C2PA's Content Credentials. The company says these watermarks will help it label content more reliably. TikTok notes that it will continue reading C2PA's Content Credentials and add them to AI-generated content made on its platform. Related to these efforts, TikTok also announced that it's launching a $2 million AI literacy fund aimed at experts, like the nonprofit Girls Who Code, to create content that teaches people about AI literacy and safety.

TikTok is letting users control how much AI content they see

TikTok is testing a new control that will let users manage how much AI-generated content appears in their video feed on the app. To help make the filter more effective, the app will also soon start adding invisible watermarks to AI content. The new AI slider will appear along with TikTok's other content controls under "Manage topics," which currently allows users to set how much they want to see from content areas including dance, current affairs, and fitness. Just like those sliders, the AI control will allow users to request to see more or less AI material, with two levels in either direction. An AI slider will only work if TikTok itself knows which content is made with AI, so it's working on improving detection too. Recognising that users often remove visual watermarks from AI-generated content, the company will start adding invisible watermarks to content made with its own tools like AI Editor Pro, along with any videos uploaded with C2PA Content Credentials -- an industry-wide authentication standard, albeit one that has already shown its limits. TikTok says both changes are rolling out over "the coming weeks," though the AI content control is only a test for now.

Less Slop Please: TikTok Will Soon Let You Reduce How Many AI Videos Appear

With over a decade of experience reporting on consumer technology, James covers mobile phones, apps, operating systems, wearables, AI, and more. TikTok believes over a billion videos uploaded to its platform were created using AI, and it'll soon launch tools to help you see fewer examples in your feed. As announced on TikTok's blog, the social media network is making changes to its AI policies to give users better control over what they see. TikTok will also improve its watermarking tech to help it better identify AI-generated content. The app says it will begin testing a new preference for AI within its Manage topics menu. It'll allow you to select whether you want to see more or less AI-generated content in your feed. The feature can be found under Settings > Content preferences > Manage topics. The brand currently has toggles for Creative Arts, Dance, Fashion & Beauty, Food & Drinks, Health & Fitness, Humor, Lifestyle, Nature, Pets, Sports, and Travel. There are five settings for each topic, including two for fewer videos, two for more, and a default level. Selecting less won't remove AI-generated videos entirely, but it will reduce how often they appear. The brand's blog post says, "The AIGC setting is intended to help people tailor the diverse range of content in their feed, rather than removing or replacing content in feeds entirely." The new toggle will be added to the top of the list, but it may be a while until it tolls out as TikTok says this is a "testing" phase and it has yet to announce an official release date. TikTok also says it will add "invisible watermarks" to content that it believes is AI-generated. The brand already uses C2PA Content Credentials tech to identify videos, using metadata to tell other platforms a video is AI-generated. As videos are reuploaded and shared by other accounts, that metadata can be lost in the process. These new watermarks are designed to follow the video no matter how many times it's uploaded.

TikTok will let users tone down the amount of AI content in their feed

TikTok is the latest social platform to hand users more control over the amount of AI-generated content they see. The app is experimenting with a new setting that will allow users to request to "see less" AI in their "For You" feed. The new toggle will be available in "the coming weeks" in the app's "manage topics" section that lets people personalize the types of videos more likely to appear in their feed. In a blog post, TikTok says the control is meant to help users "dial things down" if they want to see less AI content in their recommendations. The update follows a similar move from Pinterest, which has faced complaints about AI slop drowning out authentic images on the platform. TikTok, like most other platforms, has seen an influx of AI-generated content in recent years. The company, which requires users to label such creations, says that there are more than 1.3 billion videos with these labels. At the same time, the company notes that its existing methods for detecting AI content aren't perfect, Currently, TikTok relies on a widely-used watermarking system called Content Credentials that adds metadata to AI-generated content. But these signals can get harder to detect when images or videos are edited in other apps or copied and reshared to other sites, as often happens with viral content. To address this, TikTok says that it will also experiment with an additional "invisible watermarking" system that could help it identify and label AI content more reliably. "'Invisible watermarks' add another layer of safeguards with a robust technological 'watermark' that only we can read, making it harder for others to remove," the company explained.

TikTok has new weapons against AI for users who want it off their feed

TikTok is testing a feature to control how much AI-generated content users see on the platform. Credit: TikTok Social media users seem to be getting sick and tired of AI-generated content on their favorite platforms. And the backlash against AI-created content, or "AI slop" as it's often referred to, is growing. The big social media platforms appear to have noticed this anti-AI sentiment. Last month, for example, Pinterest launched a new feature that completely filters out AI content if users choose to do so. Now, TikTok just announced that it is soon launching its own tools to help its users spot AI-generated content and see less of it if they choose to do so. According to TikTok, the company is currently testing out a new content control in the "Manage topics" feature that will allow users to specifically adjust how much AI-generated content they see on the platform. The "Manage topics" feature already gives users the ability to tell TikTok's algorithm what type of content they want to see in the For Your feeds as it pertains to a specific content category like "Current Affairs," "Health & Fitness," or "Fashion & Beauty." The AI-generated content topic control is different then all those as it appears to be the first that spans all of TikTok's topic categories and explicitly deals with how the content is created. TikTok makes it pretty clear in its announcement that the company itself isn't anti-AI, saying it "believes AI can transform how people share their creativity, discover new passions, and stay safe on our platform, when used transparently and responsibly" and that it "invests in AI-powered experiences that create value for our community." However, the company also stresses multiple times that it's important to be transparent to users. To aid in that transparency, TikTok is also testing an "invisible watermarking" solution to help bolster its current AI label policies. TikTok says this "invisible watermark" will help "add another layer of safeguards to our current AI-generated content transparency measures." In addition to that, TikTok is also launching a new $2 million AI literacy fund and joining the non-profit Partnership on AI, which studies and develops best practices for ethical AI. TikTok's AI content controls announcement may just open the floodgates for other social media platforms to filter out AI-generated content for users who choose to do so. In fact, the question may not be will they do it but which platform is next to do it?

TikTok to give users power to reduce amount of AI content on their feeds

Platform reveals it hosts more than 1bn AI videos as it starts testing over next few weeks before global rollout TikTok is giving users the power to reduce the amount of artificial intelligence-made content on their feeds, as it revealed the platform hosts more than 1bn AI videos. The change, which is being tested over the next few weeks before a global rollout, comes as new video-generating tools such as OpenAI's Sora and Google's Veo 3 have spurred a surge in AI content online. The Guardian revealed in August that nearly one in 10 of the fastest-growing YouTube channels globally only show AI-generated videos. Many qualify as "AI slop", the term for low-quality, mass-produced content that is often nonsensical or surreal. Jade Nester, TikTok's European director of public policy for safety and privacy, said: "We know from our community that many people enjoy content made with AI tools, from digital art to science explainers, and we want to give people the power to see more or less of that, based on their own preferences." TikTok, which revealed the change at its annual European trust and safety forum in Dublin, said there are now 1.3bn videos on its platform labelled as AI-generated. Overall, more than 100m pieces of content are uploaded to TikTok every day, making AI content a small part of its overall inventory. Users will be able to reduce or increase the amount of AI content they view by opening the "manage topic" setting on the app and choosing "AI-generated content". Other topics that can already be reduced or increased under the filter include current affairs, fashion, beauty and dance. TikTok's guidelines require creators to label "realistic" AI-made content, and it bans harmful deepfakes of public figures or crisis events. Any realistic AI video that is not labelled can be taken down under its community policies. In addition, the app will now attach an "AI-made" watermark to content made with its own AI tools, or flagged by an industry-wide initiative for identifying AI-generated material called C2PA. TikTok said the move will help prevent users from trying to evade the labelling process. TikTok is also launching a $2m (£1.5m) AI literacy fund for experts and organisations such as Girls Who Code, a non-profit group teaching programming and robotics, to create educational content about using AI responsibly. The video-sharing app has defended its moderation strategy, amid controversy over plans to make hundreds of UK-based content moderators redundant. TikTok is set to cut 439 jobs in its trust and safety team in London, with trade unions and online safety experts raising concerns that humans are being replaced by AI moderation. Brie Pegum, TikTok's global head of program management for trust and safety, said human moderation would still have a role, but AI can protect employees by removing the most distressing content before it has to be seen by humans. TikTok said there had been a 76% decrease in shocking and graphic content seen by human moderators over the past year due to intervention by its automated systems. "We think that it's essential to balance humans and technology to keep the platform safe," she said. "We'll have humans as part of that process, but we also think it's very important to reduce exposure to harmful content as quickly as possible, which is where a lot of the machine support is really helping us work at speed."

TikTok will finally let you control the AI slop invading your feed

A new slider lets you decide how much AI-generated content appears What's happened? AI-made videos are everywhere now, and they aren't always easy to spot. TikTok is rolling out new tools so you can recognise them more easily and control how much of this content appears in your feed. * Clearer "Created with AI" labels and stronger automatic detection will help flag videos even when creators don't disclose their use of AI. * A new slider in Settings > Content Preferences > Manage Topics lets you increase or decrease how much AI slop you see. * A new "invisible watermark" to detect AI-created audio and video based on embedded signals, not just user tagging. * TikTok is also launching a $2 million AI-literacy fund to support educational content about responsible and transparent AI use. TikTok's been on a rollout streak, adding a Bulletin Board for creator updates and letting you share Amazon Music tracks directly in the app. Recommended Videos This is important because: Unlabeled AI clips can spread quickly and look real, especially in news-style edits, political content, and even disturbing content generated through AI apps. * The current C2PA Content Credentials can be stripped during reposts, so TikTok is testing "invisible watermarking," a hidden tag to keep AI-content identifiable even if it's edited or reuploaded elsewhere. * These watermarks will be applied to AIGC made with TikTok's own tools, including AI Editor Pro, and to uploads that already carry C2PA metadata. * TikTok will keep using C2PA Content Credentials in addition to invisible watermarks to make labeling more reliable. Why should I care? If you scroll TikTok daily, these changes directly shape what you'll see and how you understand it. * You can now adjust whether you want your feed full of AI-made content, balanced, or almost entirely human-made. It's your feed, your rules. * The invisible watermark means fewer hidden AI surprises as it adds a second, harder-to-remove layer of verification. * If you create content, know that AI-generated videos will get tagged, watermarked, and treated differently. It's smart to consider how you produce and label your work. OK, what's next? TikTok says these updates will roll out globally in phases over the coming weeks. * Expect the new labels and detection to appear first, followed by the user controls.

Inundated with slop, TikTok tests feature that will let users request to 'see less' AI generated content in their feeds

The new toggle aims to help users "dial things down" on the AI front. TikTok is currently testing a new feature that will let users request to "see less" AI generated content in their feeds, in an expansion to its AI control features which have thus far identified 1.3 billion AI generated videos posted to the platform. The social media giant announced the new tools at its annual European trust and safety forum in Dublin. In a press release (via Engadget), TikTok stated that the app's "manage topics" section will be expanded with a new toggle "in the coming weeks". This toggle will apparently help "dial things down" for people who don't want their feed clogged up with AI slop, though TikTok says the toggle is designed to "tailor" the range of videos in a feed "rather than removing or replacing content in feeds entirely." Hence, it sounds like you won't be able to opt out of seeing AI slop completely, only choose to see less of it. This toggle will be joined by a new "invisible watermarking" feature for TikTok's own AI-infused editing tools and "content uploaded with C2PA Content Credentials." TikTok already requires users who post "realistic" AI generated content to label it appropriately, in part using those C2PA Credentials which embed metadata to help identify AI generated images and videos. This is how TikTok spotted more than a billion AI vids on the app. However, TikTok says that "a common industry challenge is that these kinds of labels may be removed when content is reuploaded or edited elsewhere." The invisible watermarking tools are designed to help combat this. TikTok is careful to avoid saying anything negative about generative AI, explicitly stating that it wants to "safeguard and empower positive experiences with AI" -- whatever those are. But the general thrust of the release suggests that AI generated videos are becoming a problem for the company, or at least people using the app. According to The Guardian, more than 100 million videos are posted to TikTok daily, which means the 1.3 billion AI videos it has identified still represents only a small amount of what's being uploaded to the app (though we don't know how many videos are currently escaping detection). But alongside all these AI control functions it's adding, TikTok says it's launching an $2m AI literacy fund for organisations like Girls Who Code, which is aimed to "create For You feed content that teaches people about AI literacy and safety". That isn't something you do if people are already engaging with AI in a literate and safe manner, while everything together suggests that AI slop is becoming an image problem for TikTok. The company also recently came under fire over plans to make 439 roles in its trust and safety teams redundant, aiming to replace those moderation jobs with AI monitoring systems. The move was condemned by trade unions and online safety experts, though TikTok claimed the redundancies were part of a "reorganisation" that aimed to "evolve this critical function for the company with the benefit of technological advancements."

You can finally tell TikTok to stop showing you fake AI videos

TikTok has launched a new setting, allowing users to modify the volume of AI slop in their "For You" feed, alongside enhanced labeling technologies for such content. The AI-generated content (AIGC) control integrates into the app's "Manage Topics" tool. This capability lets users adjust content visibility, similar to existing controls for categories like Dance, Sports, and Food & Drinks. TikTok stated in a blog post that this setting aims to assist individuals in tailoring their feed's diverse content, not to remove or entirely replace content. This development occurs as companies like OpenAI and Meta increasingly adopt AI-driven feeds. In September, Meta introduced Vibes, a new feed for sharing and creating AI-generated videos. Following Meta's launch, OpenAI released Sora, a social media platform for creating and sharing AI-generated videos. Since Sora's introduction, realistic AI-generated videos have appeared on TikTok. Many TikTok users are also utilizing AI to generate visuals for posts on various subjects, including history and celebrities. TikTok indicates that users wishing to reduce AI-generated content can now do so, while those who prefer more can increase its frequency. Accessing this feature involves navigating to Settings, selecting "Content Preferences," then "Manage Topics," and adjusting the slider for AI-generated content and other topics. This change will be implemented in the coming weeks. Video: TikTok To enhance AI-generated content labeling, TikTok is testing "invisible watermarking" technology. TikTok already mandates labeling for realistic AI-generated content and employs C2PA's Content Credentials, a cross-industry technology that embeds metadata. This metadata informs TikTok and other platforms when content is AI-generated. However, TikTok noted these labels can be removed during reupload or editing on other platforms. The new "invisible watermarks" introduce an additional layer of security, detectable only by TikTok itself, making removal more difficult. TikTok will begin applying invisible watermarks to AI-generated content created with its tools, such as AI Editor Pro, and to content uploaded with C2PA's Content Credentials. The company asserts these watermarks will improve content labeling reliability. TikTok will continue to recognize C2PA's Content Credentials and apply them to AI-generated content made on its platform. Coinciding with these initiatives, TikTok announced a $2 million AI literacy fund. This fund targets experts, such as the nonprofit Girls Who Code, to develop content educating individuals on AI literacy and safety.