More producers are using AI for artistic expression, and transparency about AI in content creation is growing. TikTok stated today that it will launch a new tool to classify AI-generated video and begin testing alternative automatic labelling methods to address this concern.
AI material may confuse or mislead viewers, thus the company introduced AI labelling. Naturally, TikTok has previously amended its policy to address synthetic media, which requires users to mark AI content with realistic visuals, audio, or video, like deepfakes, to enable viewers contextualise the video and prevent misinformation. TikTok can remove unknown realistic AI images under their policies.
Except for purposely misleading users, certain AI-generated material can appear real or false. End customers prefer more transparency in this grey region so they know if the content was substantially modified or made with AI.
TikTok’s new tool will help producers comply with this synthetic media guidelines and flag other AI-generated or modified content. Creators can use the tool after posting a video, and TikTok doesn’t require them to re-label their previous videos.
When the creator uses the new tool, TikTok will post a notice below the username proclaiming the video AI-generated.
TikTok won’t penalise producers for not labelling non-synthetic media AI material.
TikTok is also developing an AI content detection and labelling system. This week, it will test a “AI-generated” label on content it determines was altered or developed with AI.
The business did not disclose how its AI content identification technology will function since it could allow bad actors to circumvent it. But TikTok said it will explore multiple AI recognition techniques and “assessing” provenance agreements to assist platforms recognise AI by inserting AI labels into content.
In recent months, OpenAI and Google announced AI detection capabilities, making AI labelling increasingly frequent for major platforms. Instagram may also be working on a feature to show AI-created or altered material. The EU wants platforms to label AI content as a rule to counter deception.
TikTok will now name any AI-using effects with “AI” as part of its transparency effort. It previously withheld that information. The Bold Glamour filter on TikTok went viral because of its ability to modify users’ appearances, although others feared it used AI instead of AR. TikTok did not reply to press enquiries about the product’s AI status.
Users may now easily identify TikTok filters that use AI with the new labelling. The revamped Effect House creator rules will ask them to do so, the business adds.
When creating its new AI labels, TikTok engaged with its Safety Advisory Councils and industry experts including MIT’s Dr. David G. Rand, who has studied how users react to different AI labels. Thus, TikTok chose “AI-generated” to be understood by all demographics.
Over the following weeks, it will release educational videos and other media literacy initiatives to help people grasp AI.
TikTok said the improvements reflect its February pledge to the Partnership on AI’s ethical Practises for Synthetic Media, a code of industry best practises for AI transparency and ethical innovation. It held roundtables with young people in August with the NGO Digital Moment to learn about their views on online AI breakthroughs.
The AI labels will roll out today, but you may not see them right away.