Amazon's threat team reported on Friday that hackers pointed autonomous AI agents at universities in East Asia. In about 11 minutes the agents found a new flaw in a medical imaging app and could write files. The app's maintainers have since released a fix.