收听本期播客
阅读正文
A recent investigation has shed light on the vast scale of content created by artificial intelligence (AI) on the social media platform TikTok, revealing that it attracts billions of views. Conducted by the non-profit organisation AI Forensics, the study uncovered that alongside harmless entertainment, a significant amount of this content consists of damaging material, including anti-immigrant stories and inappropriately sexualised images.
The report identified 354 accounts dedicated to AI content, which together had accumulated a remarkable 4.5 billion views in just one month. Researchers suggest that the creators of many of these accounts are attempting to manipulate TikTok’s algorithm, the system responsible for selecting which videos appear to users. One strategy involves posting extremely high volumes of material; for instance, one account was found to be publishing up to 70 videos per day, a rate that strongly implies the use of automated processes.
A primary concern raised by the report is the nature of the content itself. The investigation found fake news reports, cleverly designed to resemble broadcasts from reputable news organisations, which were used to promote anti-immigrant views. Another prominent category involved the sexualisation of female figures. Crucially, much of this AI material was not identified as computer-generated, which significantly increases its potential to mislead viewers. While TikTok has since taken down many of the accounts highlighted in the report, researchers noted that they had frequently been operational for several months before any action was taken.
In response to the findings, a spokesperson for TikTok dismissed the report’s claims as unsubstantiated, adding that this is an issue impacting the entire social media industry. The company also emphasised that it actively works to remove harmful content and provides tools for creators to label AI-generated material. It is worth noting, however, that not all the AI content was malicious. The most popular accounts often posted what is sometimes referred to as ‘slop’ – strange, nonsensical videos, such as animals taking part in the Olympics, designed purely to go viral.
This investigation underlines how the distinction between human and AI-created content is becoming increasingly unclear. The researchers argue that simple labels are an insufficient solution and have proposed that platforms should consider separating AI content from human-generated posts altogether.