This paper illustrates how AI risk and impact assessments may help mitigate the harms arising from specific AI systems, but it also highlights some of the limitations associated with the use of these assessments. The paper provides an overview of five AI risk and impact assessments that have been implemented or proposed by governments around the world— in Canada, New Zealand, Germany, the European Union, and San Francisco— and includes a comparative analysis of how they can help assess and mitigate risks. This paper includes analysis of both AI risk and impact assessments, which vary but are often used interchangeably, to highlight meaningful overlap and enable comparison.
Format: |
|
Topics: | |
Website: | Visit Publisher Website |
Publisher: | Center for Long-Term Cybersecurity - UC Berkeley |
Published: | July 1, 2021 |
License: | Copyrighted |
Copyright: | © 2021 CLTCBerkeley All rights reserved. |