Human rights in the age of

HUMAN RIGHTS IN THE AGE OF ARTIFICIAL INTELLIGENCE
tool to monitor their citizens, facilitate profiling of certain groups, and even identify and locate individuals.
43
Enabling discriminatory profiling: 
Facial recognition software is not just being used to surveil and 
identify, but also to target and discriminate.
44
Assisting the spread of disinformation: 
AI can be used to create and disseminate targeted propaganda, 
and that problem is compounded by AI-powered social media algorithms driven by “engagement,” which 
promote content most likely to be clicked on. Machine learning powers the data analysis social media 
companies use to create profiles of users for targeted advertising. In addition, bots disguised as real users 
further spread content outside of narrowly targeted social media circles by both sharing links to false 
sources and actively interacting with users as chatbots using natural language processing.
45
In addition, the specter of “deep fakes,” AI systems capable of creating realistic-sounding video and audio 
recordings of real people, is causing many to believe the technology will be used in the future to create forged 
videos of world leaders for malicious ends. Although it appears that deep fakes have yet to be used as part 
of real propaganda or disinformation campaigns, and the forged audio and video is still not good enough to 
seem completely human, the AI behind deep fakes continues to advance, and there is potential for sowing 
chaos, instigating conflict, and further causing a crisis of truth that should not be discounted.
46
Perpetuating bias in the job market: 
Hiring processes have long been fraught with bias and 
discrimination. In response, an entire industry has emerged that uses AI with the goal of removing human 
bias from the process. However, many products ultimately risk perpetuating the very bias they seek to 
mitigate. As in other areas a major cause of this is the prevalent use of historical data of past “successful” 
employees to train the ML models, thus naturally reproducing the bias in prior hiring.
47