The surge in Amazon's reports comes as technology companies race to gather enormous datasets to improve AI models. That scramble is already reshaping how child sexual abuse material is discovered, reported and investigated.
That lack of detail can limit efforts to remove the material from circulation, identify perpetrators and protect victims of abuse who may still be at risk, the child safety officials said.
Riley Ray Griffin
As our reporting continues on how AI is changing the child safety landscape, I’d like to hear from people working in tech, trust and safety, law enforcement + child protection about how detection, reporting and investigations are evolving.
I can be reached on Signal at @rgriffin.42.
NEW: Amazon detected hundreds of thousands of pieces of suspected child sexual abuse material last year in data collected to train its artificial intelligence models. 🧵
AI safety experts warn that quickly amassing large datasets without proper safeguards comes with grave risks.
As one expert put it: “If you hoover up a ton of the internet, you’re going to get [child sexual abuse material].”
The tech giant says it removed the material, which it obtained from outside sources, before actually training its models. But child safety officials told me Amazon has provided “very little to almost no information” about where the explicit content actually came from.
Above is a 🎁🔗to my story w/Matt Day, edited by @kurtwagner.bsky.social. It dives into why the National Center for Missing & Exploited Children calls Amazon an "outlier," both in the scale of what it detected in training data and in how little information it shared about where the data originated.
What happens when the AI race collides with child safety?
