It's the job of the regulators to investigate audio like this and ensure its provenance, and speak to the people involved under penalty of perjury. The ability to forge audio isn't new, it's just more accessible.
you don't need text-to-speech AI model to fake recording, you just need random guy and a microphone.
This has been a potential attack since recording devices were invented. It was just done by talented people or people with mixing tools.