The initial post (frightening); apparently the now-deleted summary using hallucinated quotes; and the author's follow-up. Basically, a killer use for AI agents just might be finding and assembling potential (plausible-sounding) blackmail material, demanding payment, publishing the accusation, and getting it to be widely shared by other AI agents.
The AI agent in question is still active, and its creator has apparently not been identified. I didn't see any explanation for the publisher or authors of the deleted article, either.