Researchers warn that large-language models, automated moderation systems, and algorithmic recommendation engines can trap ...