Responsible AI Agents

TL;DR

This paper discusses how to design and regulate AI Agents to ensure they act responsibly, align with human norms, and prevent harmful behaviors, emphasizing technical and legal frameworks for responsible deployment.

Contribution

It introduces a computer science approach to value-alignment and interaction discipline to improve AI Agent responsibility and argues against granting AI Agents legal personhood.

Findings

Interaction-based discipline can reduce rogue AI actions

Value-alignment improves user control over AI Agents

Legal responsibility remains with humans, not AI Agents

Abstract

Thanks to advances in large language models, a new type of software agent, the artificial intelligence (AI) agent, has entered the marketplace. Companies such as OpenAI, Google, Microsoft, and Salesforce promise their AI Agents will go from generating passive text to executing tasks. Instead of a travel itinerary, an AI Agent would book all aspects of your trip. Instead of generating text or images for social media post, an AI Agent would post the content across a host of social media outlets. The potential power of AI Agents has fueled legal scholars' fears that AI Agents will enable rogue commerce, human manipulation, rampant defamation, and intellectual property harms. These scholars are calling for regulation before AI Agents cause havoc. This Article addresses the concerns around AI Agents head on. It shows that core aspects of how one piece of software interacts with another creates ways to discipline AI Agents so that rogue, undesired actions are unlikely, perhaps more so than rules designed to govern human agents. It also develops a way to leverage the computer-science approach to value-alignment to improve a user's ability to take action to prevent or correct AI Agent operations. That approach offers and added benefit of helping AI Agents align with norms around user-AI Agent interactions. These practices will enable desired economic outcomes and mitigate perceived risks. The Article also argues that no matter how much AI Agents seem like human agents, they need not, and should not, be given legal personhood status. In short, humans are responsible for AI Agents' actions, and this Article provides a guide for how humans can build and maintain responsible AI Agents.