August 08, 2024 – On August 6th, soyacincau reported on their recent visit to TikTok’s Transparency and Accountability Center (TAC) in Singapore, shedding light on the content moderation mechanisms of the globally popular short video platform.
TikTok currently employs over 40,000 moderators worldwide to ensure platform safety, managing content in over 70 languages. These moderators, stationed in various markets, possess specific knowledge of local cultures and legal frameworks.
The transparency report reveals a significant Asian language distribution among TikTok’s moderation team. Specifically, 8.3% of moderators handle Indonesian, 4.7% Vietnamese, 3.9% Thai, 2.9% Malay, 2.8% Filipino, and 2.3% Japanese content.
According to the report, every uploaded content undergoes TikTok’s automated moderation technology, utilizing various techniques covering visual, audio, titles, descriptions, and keywords. If the system detects any violation of community standards, the content is automatically removed.
However, in ambiguous cases where the degree of violation is unclear, the automated system forwards the content to human moderators for a secondary review.
TikTok data indicates that globally, 0.9% of all uploaded videos are removed, with 97.7% of these deletions being proactive. This suggests that a substantial amount of inappropriate content is eliminated even before being reported.
In case content is mistakenly flagged as violative, users have the option to appeal. The appealed content is then revisited by moderators who decide whether to reinstate it.
Regarding fraudulent advertising, TikTok employs a three-tiered moderation approach. All advertisements undergo scrutiny based on community standards, advertising policies, and creative policies. Before an account can initiate an advertising campaign, it must undergo a vetting process to establish its credibility.
In terms of AI regulation, TikTok prohibits Artificial Intelligence Generated Content (AGC) from portraying false authoritative sources or crisis events. Additionally, the platform bans AGC impersonating any real or fictitious character on the platform.
