The two areas where this get a bit dicey are public records and scam artists.
The area where I see this getting dicey is the protection against illegal searches. Let's say XYZ data, mines all the data on liquor sales at local restaurants, from that and other sources their AI can pull a profile that links you specifically that says on a certain night of any given week at a certain hour there is a likelihood that you might test over the legal limit behind the wheel of your car. Law enforcement subscribers would know the exact place and time to check you, and you would have no protection from unreasonable searches because all the data was public. Law enforcement on the other hand is under no obligation to share the data they collect on you. A simplified example but feasible given today's technology.
I haven't started wearing a tin foil hat yet, but I see some real threats with data companies self monitoring themselves. As to the decision to "opt out" you aren't given that option when a profile is "built" from multiple sources with AI assumptions filling in the blanks.