The article already acknowledges that anyone with bad intentions will circumvent any measures like this, and any kind of mandated audio filter would likely just

The article already acknowledges that anyone with bad intentions will circumvent any measures like this, and any kind of mandated audio filter would likely just be parsed by people as a connection issue, especially over the phone with a bunch of VoIP compression already. If anything, I think requiring an actual spoken disclaimer that you're talking to an AI with an AI generated voice, like requirements most states have for disclosing calls being recorded, would make more sense.