Audio fingerprinting works because small differences in browser versions, operating systems, audio libraries, and hardware can produce slightly different results from the same generated signal. It is not necessarily enough to uniquely identify a device by itself, but it becomes much more useful when combined with canvas, WebGL, hardware, timing, and interaction data.
The gaining happens after retrieval of the results it seems: so it's just silent for the end user. And perhaps they do both fingerprinting, and ensure audio devices remain awake.
Imagine going after those seeking to protect their works from being stolen by AI, instead of the AI scrapers that caused this entire shit show to begin with. Put the blame where it makes sense, and don't try to guilt trip those that are victim of intellectual property theft, into believing they are the ones responsible for the web becoming an accessibility-nightmare; not for just the disabled, but everyone (locking content behind: a login, PoW challenges, and CAPTCHAs). Awful defeatist take.