Acting with AI: An Interaction-Based Framework for Agentic Tort Liability

In three linesA legal framework for agentic AI tort liability. The paper proposes three interaction types (autonomous drift, pure tool use, collaborative planning) and uses interaction logs as primary evidence to determine where liability attaches. Introduces a "Reasonable Agent" standard based on constraint verification and forensic logging.

Summary generated by Claude — human-verified