← Volver a análisis
Este artículo aún no está disponible en tu idioma; se muestra la versión en inglés.
IA y Derecho
AI agents: autonomy and the allocation of liability
When an AI agent can act, place orders, or call tools on its own, who is liable when it errs?
Agents that act autonomously raise new liability questions: when one places orders, sends messages, or calls external services on a user's behalf and errs, is the developer, deployer, or user responsible?
The key is the scope of authorization, controllability, and traceability — whether the user clearly granted what permissions and can stop it at any time.
Set clear permission boundaries, human confirmation for critical actions, complete action logs, and define liability and limits clearly in the terms.