Then it shows 'Export started'. After two minutes I got an email with a download link. Good one, Claude!
EDIT:
I uploaded the export of Claude to Paper Trail. Looking at the results it makes all sense in itself. It is more the way that details are grouped that makes it scary, something to think more about my AI usage.
I'd expect an entity purporting to promote privacy to provide a FOSS tool users can download to run the report themselves (those who care wouldn't mind running the report generation for a long time with limited compute), why increase the number of entities?
Wanting to be viewed as trustworthy is not the same thing as having successfully convinced potential users they are.
Comments
ChatGPT link for export: https://chatgpt.com/#settings/DataControls
After some verification I got this by email:
'We have started preparing your data export, this process may take a few days. We will email you when it is ready to download.'
So it might take a while to test the Paper Trail. I am curious though!
Claude is much easier. Link: https://claude.ai/new#settings/data-privacy-controls
Then it shows 'Export started'. After two minutes I got an email with a download link. Good one, Claude!
EDIT:
I uploaded the export of Claude to Paper Trail. Looking at the results it makes all sense in itself. It is more the way that details are grouped that makes it scary, something to think more about my AI usage.
I'd expect an entity purporting to promote privacy to provide a FOSS tool users can download to run the report themselves (those who care wouldn't mind running the report generation for a long time with limited compute), why increase the number of entities?
Wanting to be viewed as trustworthy is not the same thing as having successfully convinced potential users they are.
Good point