when you say "audio interface" I assume you mean like, "Hey Siri" and not an audio interface as in recording device, right?

So we are talking about a HomePod with a screen, or like one of those Meta "Portal" things?

Not sure. Some AI audio pendants are always on. The Apple device is rumored to adapt its interface to the user based on facial recognition. They could choose to start monitoring audio when it thinks a known human wants to interact with the device, https://news.ycombinator.com/item?id=47145201

  Apple is developing a tabletop robot as the centerpiece of its artificial intelligence strategy, with plans to launch the device in 2027.. The robot resembles an iPad mounted on a movable limb that can swivel to follow users around a room..The company is also exploring other robotics concepts, including a mobile bot with wheels similar to Amazon’s Astro, and has discussed humanoid models..