Is this personally identifiable? I don't see any issue with collecting and selling anonymized observations. It would be like police setting up cameras and using them to train a machine learning model on drunk driver detection. It's not collecting who is driving, just observing how normal cars subtly drift in and out of a lane and brake vs. intoxicated drifting/braking and using that to train a DUI detection model.
The true test is if the anonymized data can be filtered and cross-referenced in a way to still link back to people[0]. [0]: https://www.reuters.com/article/us-health-privacy/anonymous-...
Anonymous data: Red car observed on Dawn Road at 23:34 on May 31 2019. DUI predictor - likely sober. Speed - 53 mph.
Can this information be de-anonymized? Sure can!
Police: "The suspect fled 7-11 Around 23:30 and either went on Dawn Road or Lost Haven Road. Let's filter all red car observation events that happened that night between 23:30 and 23:40 on Dawn and Lost Haven. Yup, found him"
So you are saying we have to scrub it further? You end up removing all information until there is nothing left.
"car observed on May 31"