[c22]

True, but you could also use llms to autonomously engage with content you're not interested in, batch replies for times you're not around, inject coherent, consistent, plausible, but false details into your messages, or modify/flag details you didn't mean to disclose.