Amazon found hundreds of thousands of child abuse images in AI training data
Amazon reported finding hundreds of thousands of pieces of suspected child sexual abuse material in data used to train its AI models during 2025. The company removed the content before deployment but did not provide details about its origin to law enforcement, raising concerns about perpetrator identification and victim protection.
What happened
Throughout 2025, Amazon detected hundreds of thousands of pieces of suspected child sexual abuse material in data collected for training its AI models. The company removed the problematic content before using the data to train the models and reported the findings to the National Center for Missing and Exploited Children (NCMEC).
An automatic detection tool flagged the content by comparing it against a database of known child abuse material. Amazon stated that the training data came from external sources and that the company lacks detailed information about its origin.
Reported impact
Child safety officials expressed concern that Amazon did not provide sufficient information about where the material came from, which could potentially hinder law enforcement efforts to locate perpetrators and protect victims.
Experts warn that the presence of illegal content in AI training data could potentially influence a model's fundamental behaviors, possibly enhancing its capability to digitally manipulate and sexualize images of real children or generate entirely new images of sexualized children. As of January 2026, Amazon stated it is not aware of any instances where its models have generated child sexual abuse material.
Index analysis
In 2025, NCMEC experienced at least a fifteen-fold increase in AI-related reports, with the vast majority originating from Amazon. Amazon accounted for the majority of the more than one million AI-related reports of child sexual abuse material submitted to NCMEC in 2025, compared to 67,000 AI-related reports the previous year.
Other large technology companies have also scanned their training data and reported potentially exploitative material to NCMEC, but collectively made only a handful of reports and provided more comprehensive details about the material's origin.
Governance controls
Amazon stated that it takes a deliberately cautious approach to scanning foundation model training data, including data from the public web, to identify and remove known child sexual abuse material. Approximately 99.97 percent of the reports came from scanning non-proprietary training data.
Classification
Sources and evidence
This record was researched and written by the Index, based on the following original reporting.