>because screenshots are harder for computers to parse, and that's a marginal security advantage. If the user is going to send e.g. their account number to someone else (for a legitimate reason), it's better that they do it as a screenshot than that you force them to type it as text, because text is machine searchable. Which is worse when that messaging system gets compromised and then the attacker can do a text search for a pattern matching a bank routing number and be more likely to discover that message than if it was only there in a JPG.
Tbf, one could make the argument that there would have been far fewer resources dedicated to computer vision had companies made the data more accessible and had we modified PDFs to make it easier to copy test.
People will go to great lengths to bypass annoyances. Excessive false alarms is even called "alarm fatigue"
Isn't it though? You compromise some email provider so that you have read access to the emails of a million people going back 20 years. Literally many billions of emails with petabytes of attachments. Can you do text search on all of that text? Absolutely, an ordinary PC could do that in a realistic amount of time. Can you search the text within all of those images using some kind of computer vision system? Not without your own power plant.
Tbf it is 2025, not 2010, it isnt that hard