Securing the Future of Artificial Intelligence and Machine Learning at Microsoft

By Andrew Marshall, Raul Rojas, Jay Stokes and Donald Brinkman.

This document is not about AI-based attacks or even AI being leveraged by human adversaries. Instead, we focus on issues that Microsoft and industry partners will need to address to protect AI-based products and services from highly sophisticated, creative and malicious attacks, whether carried out by individual trolls or entire wolfpacks. This document focuses entirely on security engineering issues unique to the AI/ML space, but due to the expansive nature of the InfoSec domain it is understood that issues and findings discussed here will overlap to a degree with the domains of privacy and ethics. As this document highlights challenges of strategic importance to the tech industry, the target audience for this document is security engineering leadership industry-wide. Our early findings suggest that:... paper

# Conclusion The Asilomar AI Principles illustrate the complexity of delivering on AI in a fashion that consistently benefits humanity....