TikTok will seek to make hundreds of content moderators working in its trust and safety teams redundant amid a wider push towards AI moderation.
The social media platform announced on Wednesday it would look to reduce its headcount in Ireland by around 300 jobs. It told The Independent the consultation will also involve teams in other markets, but did not specify which countries will be affected.
Content moderators are responsible for keeping dangerous and harmful content off the platform. Videos on TikTok are sorted for moderation by language rather than by region, meaning cuts to moderators impact all users, rather than being concentrated on one specific region.
TikTok has said the cuts come as part of a “reorganisation to strengthen our global operating model for trust and safety” including “through the latest technological innovations”.
It comes after more than 400 staff members in TikTok’s trust and safety team left the company’s London offices last year, prompting MPs to probe the potential consequences for online safety.
open image in galleryIn November, Dame Chi Onwurah, chair of the government’s science, innovation and technology committee, said TikTok’s staffing cuts posed a “real risk to the lives of TikTok users” and questioned its ability to keep them safe. Moderators themselves have also warned cuts to trust and safety teams risk leaving users exposed to harmful online content – including deepfakes and abuse.
The platform told The Independent that 97 per cent of content removed was taken down by automated technologies between January and April this year, and that 99 per cent of content removed by AI is taken down before anyone has reported it.
But moderators previously told The Independent they do not believe AI moderation is “ready” to be rolled out at scale.
One employee said he sees TikTok’s AI get things wrong “all the time”.
open image in gallery“We have different models that are supposed to identify guns, for example,” he said. “Sometimes they do, but they make a lot of mistakes, because if you do a gun shape with your fingers, it will identify it like a gun.”
He said a model that should identify blood would often pull out “stains” or marks on a wall.
In comments sections, he said he found AI was often unable to identify “cat and mouse” games where users would communicate in certain emojis to “trick the system” and violate the platform’s community guidelines.
The news follows increasing scrutiny on social media platforms to protect children from harmful content online. Last month, Sir Keir Starmer said under-16s will be banned from using social media from as early as spring 2027. Reacting to the ban, TikTok said it “looks forward to collaborating constructively with the government on this important issue”.
A TikTok spokesperson said: “We are exploring a reorganisation to strengthen our global operating model for Trust and Safety, including proposals to evolve the way we work to ensure teams remain scalable and agile, the creation of hundreds of new specialist roles here in Dublin and redeployment opportunities, and advancing platform safety through the latest technological innovations.”