Entities in Country A provide entities in Country B with software for, say, self-driving cars. Such cars become sufficently popular in Country B. ("Sufficiently" being read as numerous enough for the next step to become important.)
Country A and Country B come into conflict. Potential causes: resources (oil, fresh water, rare earths, whatever), ideology, etc.
Country A uses its legal (and extra-legal) powers to force the entities providing said software to push out an over-the-air update of the software. When Country A flips a switch, the self-driving cars begin moving offensively, attempting to kill passengers and pedestrians and jam the road infrastructure, doing as much economic and political damage as possible. (Whoo; suppose they target specific people and infrastructure.)
It's an entertaining scenario for your favorite TV series or something, but possibly more troubling for software controlling information networks or the power grid. And the proof-of-concept of such an attack has already been shown to be successful.
P.S. We are both assuming that self-driving cars are better than human-driven cars, which fact is not currently in evidence.
- false numbers, similar with Tesla PR where they use bad statistics to prove their system is safer
- too general statistics for your situation, say car X is safer then human globally but locally in your region humans are much safer, will you send your children with car X where is probably not safer then average human(including drunks, teens and tired people)
- say car X is super safe, so you allow them to kill some people deliberately because they saved lives, so CEO of X hates people from a specific group(say gun owners, or teachers, or developers) so he decides that on 13 every month 666 people will be sacrificed