TikTok and the rules for synthetic media, content created or modified with AI

“Synthetic media that shows realistic scenes must be clearly disclosed”, i.e. “Synthetic or manipulated supports showing realistic scenes must be clearly indicated”: as stated in point 3 of the section Integrity and authenticity from the TikTok Community Guidelineswhich are the long set of rules governing the use of the popular social network and which have just been updated.

It is one of fundamental points of the update is precisely the strengthening of the rules for the management of the so-called synthetic media, which for TikTok are “content created or modified with artificial intelligence technologies”. In short, what we called deepfakes until the other day.

The rules of the perfect deepfake

That these incredible fakescreated thanks to the use of AI, would have been a problem, especially for politicians, famous people and even for journalists, this is nothing new: on Italian Tech we have often mentioned this, since since the fake Tom Cruise arrived right on TikTok, in March of 2021.

Now the app of Byte Dance is perhaps the first of the large social platforms to take such a concrete position towards this technology: “We welcome the creativity that the new artificial intelligence and other digital technologies can unlock”, and yet “AI can make it more difficult to distinguish between fact and fiction, involving both social and individual risks. For this, whoever uses it must declare it, precisely through “a sticker or a caption as synthetic, false, not real or altered”.

In any case, “we do not allow the use of public figure synthetic mediaif the content is used for sponsorship or if it violates any other policy”, the guidelines read again, and also synthetic media that “show realistic scenes not clearly marked or labeled”, which “contain the likenesses (visual or audio) of a real person, if it is used for political or commercial advocacy or if it violates any other policy” nor “material that has been edited, merged or combined in such a way as to mislead people about real events.

Clips made this way will also follow the moderation process of all other content published on TikTok: “Videos are initially reviewed by our automatic moderation technology (which is based on artificial intelligence, ed), which aims to identify content that violates our Guidelines” and “if a potential violation is detected, the automated moderation automatically removes it”. Because it takes an AI to catch another AIObviously.

Cybercrime

The new frontier of telephone scams: the voices of relatives replicated with AI

by Francesco Marino

08 March 2023

TikTok and the “4 Pillars of Moderation”

That’s not all, because in the rules update, which would be “the most extensive so far” and is in force starting April 21, there are also more attention to the integrity of electoral consultations (explained here in detail) and a new approach to the accounts of governments and politicians, the addition of the word “tribe” among those not allowed when talking about racism and hate speech and also a reorganization by subject areas, with the beginning of each summarizes more clearly what is allowed and what is not.

TikTok also highlighted what the company calls “the 4 pillars of our moderation approach”: remove content that violates policies, limit only for over 18s access to content that requires greater maturity of judgement; exclude unsuitable content from the For You pages to the whole public; give the community greater control over the user experience, with tools and information resources.

@capoema