Risk management in Meta’s news moderation policies on Instagram

Grounded within the framework of risk management theory, this study examines the moderation polices followed by Instagram before its decision to cancel third-party fact-checking practices in early 2025. More than 660,000 Instagram news items posted by 40 news channels were collected in October 2024. In total, there were 1,219 moderated news items based on either having misinformation or being sensitive, such as containing graphic content. The majority of news items were either automatically or manually flagged as being sensitive, and only 93 posts were labeled as having misinformation. Then, manual analysis of moderated posts was conducted to classify them based on 18 different topics, and another analysis was conducted to assess the accuracy of fact-check posts. The results show that news on Palestine is the most moderated topic, as Meta flagged it as sensitive. This topic is followed by internal violence and then accidents. Out of all the channels we examined, Palestine TV is the most frequently moderated outlet, followed by Al Jazeera Arabic, TRT World, and Al Jazeera English. In terms of misinformation labels, there were some false positives, indicating clear problems in automated fact-checking. The discussion is supported by internal documents leaked by the Facebook whistleblower, Frances Haugen, to better understand the inner workings of Meta’s moderation. We argue that Meta has developed and enforced several algorithms to mitigate risk that would help in putting less government and public pressure on the social media site, which could ultimately also decrease legal, economic, and ethical liability.