The White House published a draft statement of rules for how AIs treat people. Although these rules are not legally binding, it makes sense to test a set of rules — for a while — before making it binding.

However, from what I can see in this article, these rules are based on the usual wrong assumption: that rules about the use of data about individuals can eliminate the danger posed by of accumulation of that data.

Does it require an AI used to judge people be free software? If not, we know it is too weak.

The dangers of collecting data about you include repression, for instance when China does it, and putting you at a disadvantage, for instance when a John Deere tractor does it.

To make data collection by tractors respect the rights of farmers, it must deliver that data about a given farm only to the farm owner, and if an AI analyzes the data, that should be free software running on the farmer's computer.