Jakarta - Amazon confirmed that it had found child pornography content (child sexual abuse material/CSAM) while reviewing artificial intelligence (AI) training data from external sources. The findings were reported to the National Center for Missing and Exploited Children (NCMEC) in the United States.
Citing a Bloomberg report, NCMEC received more than one million reports related to AI-based CSAM throughout 2025, with most of the reports coming from Amazon. The company stated that the content was detected during the process of scanning data from third parties that were used to train their AI systems.
However, NCMEC considers Amazon's report to be prominent because of the lack of source information. NCMEC Executive Director Fallaon McNuty said reports from technology companies generally include enough details for law enforcement to follow up. In the case of Amazon, the absence of information on the origin of the content makes the report difficult to use for investigations.
According to McNuty, this condition raises serious questions about the process of collecting data for AI training and confirms the need for stricter oversight standards and security mechanisms across all technology platforms.
In response to the criticism, Amazon stated that it used extensive scanning rules when reviewing training data, which could result in a large number of false positives. Amazon also confirmed that it had found no evidence that CSAM was actually used in the training of their AI models.
The company explained that the report was submitted through a separate system because the data came from a third party and did not include user information, so the source details could not be attached.
Amazon reiterated its commitment to responsible AI development and child safety protection. Ethical issues and security in AI development itself continue to be in the spotlight, as a number of lawsuits against other technology companies such as OpenAI, Character.AI, and Meta related to data usage and user protection.
The English, Chinese, Japanese, Arabic, and French versions are automatically generated by the AI. So there may still be inaccuracies in translating, please always see Indonesian as our main language. (system supported by DigitalSiber.id)
