Hello everyone,
I've been trying to gather some answers to your questions.
This feature has been completed rolled out at the end of last year.
PDF files are generated by many different applications which has consequences for how those documents are made searchable. Even though as an end user, it appears that a PDF is one format, how the PDF is created makes a big difference in how to make it searchable. In SharePoint there is already a search function makes many types of PDFs searchable. There's no plans currently to extend the work of the image recognition team to PDFs imminently but engineering is aware that this is a concern, but there are many nuances to how to make this cover every situation.
The data extracted is processed and lives wherever the data is stored, which includes geo support for data sovereignty.
Hope this helps.
Microsoft