Researchers show how AI programs can be sabotaged by even subtle tweaks to the data used to train them.

Read more from our friends at Wired.com