Register, C., Khan, M. A., et al. (2024).
Pre-print.
Abstract
Relationships between humans and artificial intelligence (AI) raise new concerns about privacy. AI raises new threats to privacy as it becomes more like humans in language and appearance, more observant, and more inferentially powerful. As humans increasingly form relationships with AI, we expose ourselves in new ways to technology that we don’t fully understand. Further, if AI is given the capacity for some type of awareness, it may be able to infringe privacy in radically new ways. Drawing from recent empirical work in psychology and from the contextual integrity theory of privacy, this article analyzes some of the ways that human-AI relationships may threaten values that privacy functions to promote. We then propose six tentative policies to guide the design and development of AI products to mitigate these threats to privacy.
Here are some thoughts:
The increasing integration of artificial intelligence (AI) into daily life raises significant concerns regarding privacy, particularly as AI becomes more human-like in language and appearance. As humans form relationships with AI, they expose themselves in new and often unintended ways to technology that remains complex and not fully understood. The potential for AI to possess some form of awareness introduces the possibility of radically new privacy infringements. Drawing from recent empirical research in psychology and the contextual integrity theory of privacy, this analysis explores how these human-AI relationships may threaten the fundamental values that privacy aims to protect.
To address these emerging threats, we propose six tentative policies aimed at guiding the design and development of AI products to better safeguard privacy. Currently, AI systems are already prevalent in our lives, collecting vast amounts of information, and their observational capabilities are expected to expand further. This growing presence of observational AI poses a significant risk to privacy, which is likely to intensify as relationships between humans and AI deepen. The specific implications for privacy ethics remain challenging to predict, influenced by both technological advancements and the nuanced psychology underlying human-AI interactions. While we have outlined various ways in which AI may impact privacy and suggested policies to mitigate potential harms, there is still much work to be done to foresee and adequately address the privacy risks that lie ahead. Balancing the progress of AI with robust privacy protections will be crucial in navigating this evolving landscape.
