The difference between (US) federally-funded human subjects research and companies collecting is that human subjects need to give their informed consent, except in a few situations where they would be non-identifiable. Companies, TV shows, and the like routinely get away with things that would be unethical if it were truly research.
Requiring back-reporting to participants seems reasonable, but this would probably actually be a bigger threat to identity theft, and a huge burden of additional red tape. A lot of human subjects research goes on with anonymous participants. Once you start collecting contact information, you need to store this in a secured manner, have a policy for destroying it, etc. Tying identity to the data makes for medical-record-type security requirements, even for things like visual priming and social psychology tests. Would you really want your behavior on a test (maybe of implicit attitudes toward racist sterotypes) tied to your identity in a database 'secured' on a graduate student's laptop? Thousands of such data sets exist, but currently it is usually anonymized so there is no contact information or identifying data. The results of such scientific tests, if tied to identity, can usually be compelled to be released with a court order unless the experimenter makes special arrangements upfront. Do you want your genetic screening done when you were a college freshman subpeonaed in divorce court?
The need to make individual reports is mindboggling complicated, knowing the typical technical abilities of most researchers in behavioral and biomedical areas. Merging data and finding mean differences between groups is hard enough; taking those back to individual results, and producing dozens or hundreds of individual reports for participants who mostly don't care is a huge waste of resources. You would need professional data handlers on many studies, and it would really eliminate the feasibility of a lot of small-scale and pilot studies that are not funded by large grants, studies people run as an undergrad or in graduate school as they learn about a topic. I'd guess that more than half of all studies conducted on humans go unpublished, many of them of this nature. If a researcher deems the results not worth reporting, or not being likely to make it through journal review because it is not 'surprising' enough, or if the results reveal a confound that needs to be eliminated in a further study, would you still report back to the individual? Sharing data that has not undergone peer review could be a big mistake.
Finally, even highly unreliable measures can be used to understand differences if you collect enough data. Giving specific results for unreliable measures, or giving people results if they were in a 'control' group, would be useless and might be unethical in some cases.