| Attachment | Size |
|---|---|
| AI Safety_6.pdf | 1.95 MB |
Abstract
ACT-IAC White Paper: Safety, Privacy, Security, in an Artificial Intelligence-Empowered World
Developed by the Emerging Technology Community of Interest, Artificial Intelligence Working Group
Published August 9, 2022
This paper is intended to provide a broad overview, analysis, and framework for addressing the critical elements of safety, privacy, and security in artificial intelligence (AI). Building trustworthy and ethical AI increasingly requires that elements be clearly defined, measurable, and that roles and responsibilities for their custodianship are delineated. While safety, privacy, and security cannot be guaranteed, risks can be managed. Safety: AI systems should be designed, developed, and deployed with safety in mind. Education should illuminate the indicators that could potentially harm, hurt, or hospitalize the public. Rigorous testing needs to validate and evaluate the potential for harm. Privacy: AI systems should provide adequate education to data providers outlining their privacy risks associated with data-collecting devices. Systems should be monitored to protect privacy based on policies and standards. Security: AI systems should continuously benchmark the system’s dependability over its lifespan to ensure they are protected against attack. It is incumbent upon everyone involved to monitor and measure the performance of these technologies to ensure they are serving us in the way they were prescribed.
