To Their Credit

Video

TikTok expands AI labels and election safeguards in Nigeria

The platform removed 4.38 million videos in Nigeria in one quarter as it prepared for the Jan. 16, 2027 election.

Ledger: ResistedAI is held back

TikTok has expanded its election-integrity safeguards in Nigeria before the country’s general election on Jan. 16, 2027, combining stricter treatment of political accounts with AI-content labels, moderation measures and an in-app source of voting information.

The platform removed 4.38 million Nigerian videos for violating its Community Guidelines between April and June. Of those removals, more than 210,000 concerned civic and election integrity, misinformation, or harmfully misleading AI-generated content. More than 99% of the latter group was removed before users reported it.

Across all Nigerian removals in the quarter, 98% were taken down before a user report and 91.7% were removed within 24 hours. TikTok said the enforcement was intended to address online hate, information manipulation and deceptive behaviour that could erode confidence in elections and democratic processes.

The measures were presented at a TikTok Nigeria Election Media Workshop in Lagos, held with the Nigeria Union of Journalists. Code for Africa gave participants practical instruction on identifying, checking and escalating potentially harmful or misleading material, while the sessions covered election integrity, AI-generated content and media literacy.

TikTok said it was also training more than 300 media professionals in Lagos, Abuja, Ogun and Oyo. Its preparations include an election task force with expertise in platform integrity, cybersecurity, deceptive behaviour, misinformation and elections, working with moderation technology and local-language moderators.

The company has launched an in-app Election Centre with the Independent National Electoral Commission of Nigeria and the fact-checking organisation DUBAWA. The centre provides voting information, election dates and advice on recognising misinformation, and directs users to an AI media-literacy hub explaining how to identify AI-generated material and use AI responsibly.

TikTok said it removes election misinformation including false claims about voting procedures and accounts impersonating politicians or news organisations. Claims that cannot immediately be verified may be given an unverified-information label, accompanied by a warning before sharing and excluded from recommendation in the For You feed.

Its AI measures require synthetic content to be labelled, using a creator toggle, C2PA Content Credentials and invisible watermarking to support identification. TikTok said more than 5 billion videos had been labelled through its AI-labeling tools since it introduced the AI-generated-content label. A separate account of the Lagos workshop attributed a lower figure of more than 3 billion AI-generated-video labels to Keaglie Makgoba.

Political advertising is prohibited on TikTok, including paid creator promotions for politicians and political parties. Government, politician and political-party accounts cannot buy ads, face additional requirements including two-step verification, and are restricted from monetisation features. Election-management bodies may advertise essential voter information on registration, polling and participation.

The company said it works with more than 20 fact-checking organisations accredited by the International Fact-Checking Network worldwide, as well as Code for Africa in Nigeria, media-literacy initiatives and law-enforcement agencies. It said its Nigerian preparations drew on work securing more than 300 elections globally.

The new safeguards arrive amid wider concern about AI-enabled deception in Nigerian politics. Deepfakes falsely linking candidates to terror-related activity circulated before the 2023 election, according to Deutsche Welle. INEC chairman Joash Amupitan said the commission was investigating apparent discrepancies in the preliminary voter register that may have been created through AI use, and INEC has announced a dedicated unit to monitor AI threats and detect deepfakes.

Nigeria’s 2023 election had a turnout of 27%, the lowest since the country returned to democratic rule in 1999. The Centre for Democracy and Development documented TikTok’s growing prominence during that campaign, as politicians used it to reach new audiences, and found that election disinformation had included false impersonation, manipulated audio and video, hashtag manipulation and coordinated networks.

The number of active social-media users in Nigeria had risen from 27 million in 2019 to 36 million before the 2023 elections, according to the centre. Its assessment also warned that social-media companies had often failed to respond to, label or remove reports of political misinformation because of staffing constraints.

More from Video

All Video news
  1. Pika previews Enact, an AI video tool built for live directionLedger: Adopted
  2. Seoul AI Content Festival pairs generated media with dealmakingLedger: Adopted
  3. Zheng Yuanjie turns his fairy tales into free AI animation on DouyinLedger: Adopted