Depending on the source DNA seems to have a half life of 500-1500 years. This source calculates with slightly over 500 years. Permafrost samples may be more on the upper end of the scale but the message would be similar:

A study of DNA extracted from the leg bones of extinct moa birds in New Zealand found that the half-life of DNA is 521 years. So every 1,000 years, 75 per cent of the genetic information is lost. After 6.8 million years, every single base pair is gone.

https://www.sciencefocus.com/the-human-body/how-long-does-dna-last

If you say it like that, I’ll happily abandon science to go velocitaptor jousting with you!

Even under the best preservation conditions, there is an upper boundary of 0.4–1.5 million years for a sample to contain sufficient DNA for sequencing technologies.

https://en.wikipedia.org/wiki/Ancient_DNA

No, that’s ridiculous.

This Regulation does not apply to the processing of personal data: […] © by a natural person in the course of a purely personal or household activity;

As is stated, the call is processed locally in the user’s device. If that holds true, there is no recording and no third party processing going on. Your point does not make sense.

That’s a real world issue. AIs training on each other’s output and devolving because of it. There will be a point when vendors infringing on user content and training their AIs with it will leave them worse off.

It’s easy to train a model to do exactly what you want and have the seeming “personality” that you want. It’s just incredibly expensive. You need to vet and filter everything that you use to train the model. That’s a lot of person hours, days, years. The only reason the models act the way they do is because of the data that went in to train them. If you try and fit the model after the fact, it will always be imperfect and more or less easy to break out of those restrictions.