Officers from the US Internet Crimes Against Children (ICAC) taskforce have alleged that artificial intelligence systems used by Meta to detect child sexual abuse material are generating thousands of low-quality reports, placing immense strain on law enforcement resources.
The concerns emerged during testimony in a case brought by New Mexico’s Attorney General against Meta. Investigators claimed that many of the tips forwarded to authorities lack actionable evidence or do not amount to criminal activity.
AI reports double, quality questioned
Under US law, online platforms must report suspected child sexual abuse material to the National Center for Missing & Exploited Children (NCMEC), which then routes tips to enforcement agencies.
Meta accounted for 13.8 million reports in 2024, forming the majority of 20.5 million tips received nationwide. However, ICAC officers stated that the number of cybertips their department received doubled between 2024 and 2025, with many deemed “unviable”.
Officers suggested that expanded reporting obligations under the REPORT Act may have contributed to the surge, alongside increased reliance on AI detection tools.
Law enforcement stretched thin
Investigators emphasised that every tip must be reviewed, even if incomplete. Missing images, redacted material or context gaps often prevent further action.
One officer described morale as being affected by the overwhelming volume. “We are drowning in tips,” he said, noting limited manpower to keep pace.
Meta has defended its approach, stating it cooperates closely with law enforcement and uses both AI and specialist safety teams. The company also cited legal requirements, including court rulings that necessitate warrants in some cases.
The debate reflects the broader challenge of balancing aggressive detection of online child exploitation with ensuring quality, actionable intelligence for investigators working to protect vulnerable children.