A TikTok safety feature trusted by millions of parents is not preventing children from encountering graphic videos that portray rape and torture, The Mail on Sunday can reveal.
The platform’s “restricted mode” is designed to screen out harmful content, yet apparent loopholes are enabling companies to profit by pushing disturbing AI-generated videos at underage users.
The clips appear to take advantage of weaknesses in TikTok’s content policies, which prohibit “realistic violence” and “sexually suggestive content” but can leave room for animated material to circulate.
Among the firms drawing scrutiny is the start-up Director AI, whose video-making tools have reportedly helped users generate more than 15,000 unsettling clips, many shaped by algorithms to maximise engagement.
One video posted across its network of around 50 accounts has been viewed more than 120 million times. It shows a cartoon doctor using a drill to gouge a patient’s face during a grotesque operation intended to make him resemble Kim Kardashian.
Another shows a tattooed burglar exacting revenge on a wealthy woman who once mocked his poverty, ransacking her mansion.
Footage cuts to his muscular hand clamping around her wrist while she lets out a bloodcurdling scream.
He is then seen laughing manically as he walks away with the zip of his shorts undone, carrying a holdall of stolen cash and gold bullion, leaving her battered, bruised and weeping.
A TikTok safety setting relied upon by millions of parents is failing to stop children being shown graphic videos depicting rape and torture, The Mail on Sunday can reveal
The videos exploit gaps in TikTok’s rules, which ban ‘realistic violence’ and ‘sexually suggestive content’ but allow animated depictions to slip through
Last night, after seeing the footage, Shadow Education Secretary Laura Trott said: ‘We would never tolerate children being exposed to this content offline, so the same must apply online.
‘Social media companies should not have access to our children until they can prove their platforms are safe. Labour must stop dragging their feet and get children off social media.’
For our investigation, we set up a TikTok account on ‘restricted mode’, with the age set to 14 years old, and watched hours of content about children’s TV programmes.
We put limitations on profanity, firearms and weapons, sexually suggestive content, drugs, violent or threatening imagery, and mature or complex content.
Despite this, the account’s main feed was bombarded with a significant volume of content within each of these categories.
The firm behind the videos is run by former Snapchat director Ben Chiang. Mr Chiang – a married father of one – has openly referred to the output as ‘AI slop’. Earlier this year, he boasted on LinkedIn that the videos have racked up more than 10billion views.
Those who want to keep watching must pay for access via the firm’s companion app, Reel AI, rated on Google Play as suitable for children as young as seven.
It offers up to ten free episodes of each mini-series before charging viewers per episode. One purchased episode shows a mother placing her baby in a bathtub and using a hairdryer to electrocute it.
RELATED: UK’s AI Superpower Push Risks Draining Water Supplies
Online safety expert and psychotherapist Dr Catherine Knibbs said: ‘This is about creating wealth at the expense of children.
‘You have to be a sincerely disconnected human being to not see the harm that this meticulously curated content can cause.’
One video shows a tattooed burglar exacting revenge on a wealthy woman who once mocked his poverty, ransacking her mansion
Mr Chiang co-founded Director AI last year with two former colleagues. When asked who owned the TikTok accounts that made the videos, Mr Chiang, who sits on the board of a youth homelessness charity, told the MoS: ‘It is inaccurate to characterise Director AI itself as producing and publishing all of the videos made using our tools.
‘Director AI provides AI video-production tools that creators use to make content. Those creators make their own editorial decisions and choose where to publish their work, including on third-party platforms such as TikTok… They may choose to promote Reel AI, but we do not own their accounts or pay them specifically to promote the app.’
Mr Chiang added: ‘Some content created using our tools may contain fictional mature themes, but TikTok and other platforms themselves determine eligibility for and filtering within their Restricted Mode under their own systems and policies.’
Following contact by the MoS, the age rating for Reel AI has been changed ahead of the next update to the app.
A TikTok spokesman insisted the platform’s community guidelines prohibit content promoting or glorifying violence and sexual abuse or exploitation.
They said any videos violating the standards have been taken down. Following contact by the MoS, TikTok removed the videos cited in this report – but all the accounts that made them are still active, and still show other similarly inappropriate material.
A spokesman for the Department for Culture, Media and Sport said that broadcast regulator Ofcom has the Government’s ‘full backing’ to ensure tech companies comply with the Online Safety Act.
‘Those firms have a legal duty to remove illegal content, as well as protecting children from harmful material,’ they said.
Meta buried child-safety risks
Instagram owner Meta failed to publicly disclose damning internal findings exposing ‘serious issues’ putting children at risk on the service, The Mail on Sunday can reveal.
A report from December 2022 – buried in court documents provided as part of a deposition by Sir Nick Clegg in the case brought by 29 US states against Meta – contained ‘high-risk findings’ over child safety.
It revealed that on one sampled day, around 1.4million accounts that Instagram had itself flagged as potential child-safety risks were recommended to teenagers.
On another, 924,000 potential ‘violators’ were shown the account of at least one unconnected teenager.
These were unverified adult accounts that had either received at least one child sexual exploitation violation in the previous 90 days or been flagged by Meta’s automated system as potentially posing a risk of inappropriate interactions with children.
Sir Nick, who was Meta’s then president of global affairs, said he forwarded the audit of Instagram’s youth safeguards to Meta’s legal team but did not know what action was taken.