Meta's AI Floods Cops with Useless Child Abuse Reports

US Internet Crimes Against Children (ICAC) taskforce officers state that Meta's artificial intelligence software is producing a significant volume of unviable reports about child sexual abuse. These reports are described as "junk," overwhelming investigators and hindering their ability to pursue genuine cases. Thousands of these low-quality tips arrive monthly, doubling the department's intake from 2024 to 2025.

These unviable tips, originating from platforms like Instagram, Facebook, and WhatsApp, sometimes lack criminal indicators or essential evidence such as images or text. Officers express frustration, noting that while they know crimes occurred, the lack of information prevents perpetrator identification. A Meta spokesperson, however, stated the company supports law enforcement and has improved its tip reporting process, citing swift cooperation with authorities.

Internal Meta documents from early 2019 revealed executive concerns about policing child sexual abuse, particularly with the planned end-to-end encryption in Facebook Messenger. Executives worried that encryption would prevent the detection of child exploitation. Meta now asserts that its developed safety features are designed to function even within encrypted chats.

Social media companies are legally mandated to report detected child sexual abuse material to the National Center for Missing & Exploited Children (NCMEC). Meta is the largest contributor to NCMEC, submitting millions of reports annually. The recent surge in unviable tips from Meta has been linked to the implementation of the Report Act in November 2024, potentially indicating the company's effort to comply with broader reporting obligations, though many tips now appear AI-generated and irrelevant.