Abstract
<title>Abstract</title> <p>Anthropomorphic AI systems pose two distinct ethical challenges: deceptive anthropomorphism that exploits cognitive vulnerabilities, and a responsibility gap that emerges when these systems cause harm. Existing governance frameworks adopt a developer-centric perspective, focusing on optimizing user engagement rather than enabling public accountability. This paper advances two constructive contributions to bridge the divide between developer-centric and stakeholder-centric approaches in AI ethics. First, it proposes the “tool-owner” relationship as a potential foundational principle for AI governance. The principle is simple: AI is a tool having no agency and not a human replacement; accountability flows between producer and owner through a framework that bridges their respective responsibilities. Developers may design highly anthropomorphic AI tools, provided they uphold the principle of integrity and safety. Any failure triggers full moral and legal accountability. Second, it introduces the Internalized Externality Accountability Framework (IEAF) as the operational bridge between the tool-owner principle and measurable accountability. The IEAF is a model of three tiers (Computational Tool, Psychological Influencer, and Unavoidable Parasocial Attachment), which based on two standards: (1) whether the product is designed to induce emotional engagement or trust, and (2) whether the user can voluntarily uninstall. The framework functions as a sword of Damocles and developers are granted a window to advance their products based on their disclosures and public trust, but the reverse-pressure mechanism continuously reminds them of their promised safety and integrity obligations. If a victim emerges, the reliability of the developer’s disclosure determines their accountability; if the user was highly controlled by the developer’s algorithm, the developer may face retroactive escalation to Tier III. Drawing on economic theories of externalities, psychological research on parasocial attachment, legal doctrines of social failure and undue influence, and the Resonant Amplification Framework, the IEAF is operationalised through validated instruments and evaluated via three case studies. It describes a stakeholder-centric lens that enables regulators, courts, and the public to assess developer accountability without technical expertise.</p>