Trust among AI Agents, Cybernetic Avatars and Natural Persons

Abstract

AI Agents or Cybernetic Avatars (CAs), which act as surrogates for natural persons on the Internet, are beginning to gain importance as an important application of AI technology in society. These can be positioned as components of the metaverse. Trusts between natural persons themselves and AI Agents or CAs, and between AI Agents or CAs and other natural persons or software with which they interact via the Internet, are necessary for smooth interactions among them. In this report, we discuss this trust from the perspective of its legal status and technical feasibility. Since there is a wide range of research on trust, we will introduce an overview and studies related to the content of this paper. Next, studies on the legal status of CAs will be introduced. The concept of ID authentication, which is the basis of Trust, is then explained. Next, assuming (1) natural persons themselves, (2) AI Agents and CAs as their proxies, and (3) natural persons, AI Agents, and CAs that interact with them via the Internet, we discuss the framework of ID authentication and trust that should be established among (1), (2), and (3), and the issues that arise in relation to individual ID authentication and trust. (1), (2), and (3). In particular, we will discuss hijacking and spoofing by malware and BOTs, and how to deal with them. Next, we describe the problems when one person controls multiple AI Agents or CAs, and when multiple people control a single AI Agent or CA.