Deanonymizing online accounts isn't new. Speech patterns & unique combinations of identifiers have been able to do it for a while. What's different now is that AI can do this at scale, and its getting better at it. What's also true is that most people are underestimating the danger they are in.

If you don't fear being identified and monitored by the government via Palantir (you should), then you should at least fear cyber-attackers and criminals being able to do the same. If you think the latter sounds far-fetched, consider that Big Tech is insisting AI has no boundaries or regulations. If you don't think criminals won't take advantage of that situation, then you're a fool.

Research - Large-scale online deanonymization with LLMs, 24 pages PDF