Affiliate Ram Shankar Siva Kumar and coauthors suggest ways that Microsoft and other tech giants can mitigate the security risks inherent in emerging AI technologies.