Violent extremists are applying AI to evade social media moderation, according to the leader of one of Europe’s leading violence prevention tinquire forces.
Speaking at TikTok’s Trust and Safety forum in Dublin, Judy Korn, managing director of the Violence Prevention Network, stated the nature of extremism was modifying along with the tactics applyd.
“[While] Islamist and far-right extremism is rising,” stated Ms Korn, “the sharpest rise in ideologies is in unclear violence and nihilistic violent extremism.”
Nihilistic extremism, she stated, isn’t driven by an ideology beyond simply “destruction and chaos” – which builds it harder to catch on social media becaapply it won’t trigger traditional content filters.
At the same time, violent extremists are applying generative AI to spread their messages.
Please apply Chrome browser for a more accessible video player
Musk, X and the British right-wing
“Generative AI is much more clever than violent extremists, becaapply it learns quicker than a human being how to produce content that conveys the desired message without violating regulations and without violating the platform guidelines,” stated Ms Korn.
Violent extremists are also receiveting younger.
In the UK, more than 50% of people referred to the deradicalisation programme Prevent are under 18, and one in five people arrested for terrorism is legally a child, according to counterterrorism police.
During the event, TikTok announced a number of new measures to better protect its applyrs, including new educational prompts for applyrs searching for terms related to extremism in Germany.
Read more from Sky News:
Paris’s Louvre closes gallery weeks after heist
Hacker ordered to pay back over £4m in Bitcoin
As well as measures around extremism, TikTok also announced a new invisible watermark that will support applyrs recognise when content has been built with AI.
The company has had AI detection tools for years, including the widely adopted C2P labelling system, which embeds information into the metadata of content.
“The common indusattempt challenge is that with these methods, the labels might receive reshiftd if you download the content re-uploaded or re-edited elsewhere,” stated Jade Nester, director of data public policy, Europe at TikTok.
“These invisible watermarks support us address this by adding a robust technological watermark that only we can read.”
TikTok also announced the rollout of a new, gamified “wellness hub” to encourage applyrs to focus on their mental wellbeing through meditation, affirmations and controlled screen time.
Leave a Reply