It's pretty unlikely that Meta is actually eavesdropping on your conversations, because it'd be immediately obvious from battery usage. The ability to turn speakers into microphones doesn't help if the speakers aren't actually connected to an ADC, and both of the modern smartphone OSes limit you to on the order of hundreds of samples per second, so it's rather difficult to get anything sensible without either doing a bunch of local analysis or exfiltrating it, both of which would be visible.

It can be done with neural networks [1]. Also, speech doesn’t need much bandwidth to be intelligible. You would need control of the analog filter between the accelerometer and the ADC. With 250/s acceleration samples you can reconstruct a signal of a bandwidth of more than 100 Hz anywhere in the spectrum. That is called undersampling.

[1] https://dl.acm.org/doi/abs/10.1145/3478102