Child abuse images removed from AI image-generator training source, researchers say

Researchers have discovered that child sexual abuse material was embedded in a widely used dataset for training artificial intelligence image generators, raising urgent questions about how such toxic data escaped detection and what other poisoned datasets may be quietly powering the AI systems millions rely on daily. The finding suggests that major tech companies' vetting processes for training data remain dangerously inadequate, even as these systems become increasingly central to commerce and communication. Until now, the AI industry has largely operated in the shadows on this front, with little public accountability for how it sources and screens the material that teaches machines to generate images.

yourcentralvalley.com·Aug 30