As more creators turn to artificial intelligence for artistic expression, there is also a broader push for transparency in the involvement of artificial intelligence in content creation. To solve this problem, TikTok announced today that it will launch a new tool that allows creators to annotate AI-generated content, and will begin testing other methods of automatically annotating AI-generated content.
The company said it believed the introduction of AI tags was necessary because AI content could confuse or mislead viewers. Of course, TikTok has updated its policy on synthetic media, requiring people to label artificial intelligence content containing realistic images, audio or video (such as deep fakes) to help viewers clarify the ins and outs of the video and prevent the spread of misleading information. TikTok's policies allow it to remove undisclosed images of lifelike artificial intelligence.
However, aside from extreme cases where AI is used to intentionally mislead users, some AI-generated content may waver between appearing authentic or false. In this gray area, end users will generally appreciate more transparency so they can tell whether the content they are viewing has been heavily edited or created using artificial intelligence.
TikTok is now launching a new tool that will not only make it easier for creators to comply with existing policies on synthetic media, but also flag any other content that is generated entirely using AI or has been heavily edited by AI. Creators can use the tool when they upload a video, and TikTok doesn't want creators to go back and re-tag their past videos.
When a creator uses the new tool, TikTok will display a message under the video's username stating that the creator has marked the video as AI-generated. However, TikTok will not penalize creators for failing to label artificial intelligence content that does not fall under existing synthetic media policies.
In addition, TikTok also said that it is developing a method to automatically detect and label artificial intelligence content. This week, TikTok will begin testing an "AI generated" label that will eventually apply to any content detected as being edited or created using AI.
The company declined to reveal specific details about how its technology will look for possible AI content, noting that sharing those details could allow bad actors to bypass its detection capabilities. TikTok did say, however, that it would be testing different AI detection models and was "evaluating" a provenance partnership aimed at helping the platform better detect AI by embedding AI tags into the content itself.
In recent months, both OpenAI and Google have announced their own AI detection capabilities, and labeling AI is becoming a common practice among large platforms. Instagram also appears to be working on a feature to highlight content created or edited using artificial intelligence. The European Union is pushing for platforms to label artificial intelligence content as a general rule to combat disinformation.
To increase transparency, TikTok will now also rename all special effects that use artificial intelligence to clearly indicate "AI" in the name. This is something it has previously refused to disclose. In fact, when TikTok's "BoldGlamor" filter became popular for its superb technology in changing users' faces, some people suspected that the filter used not AR (augmented reality technology), but AI. However, TikTok refused to respond to media inquiries at the time about whether the product was artificial intelligence.
With the new label change, users can now determine at a glance which TikTok filters use AI. The company said its updated Guidelines for Effect House Creators will require them to do the same.
TikTok noted that when developing new AI tags, it consulted with its safety advisory board and industry experts including Dr. David G. Rand of the Massachusetts Institute of Technology, whose research examined user reactions to different types of AI tags. To this end, TikTok hopes that the term "artificial intelligence generation" can be widely understood by different groups of people.
In the coming weeks, TikTok will also launch educational videos and other media literacy resources to help users better understand artificial intelligence.
These updates follow TikTok's other initiatives around artificial intelligence in media, including a February commitment to join the Partnership on AI's Responsible Practices for Synthetic Media, an industry code of best practices for transparency and responsible innovation in artificial intelligence. In August this year, the company also partnered with the non-profit organization Digital Moment to host a roundtable with young people to understand their views on the online development of artificial intelligence.
AI tags are rolling out starting today, but you may not see them right away until the rollout is complete.