Agentic AI systems—AI that can act autonomously—are increasingly used as personal assistants and virtual employees. A new preprint on arXiv argues that these systems are likely to become common subjects of digital investigations and incident response. Yet the paper notes that little is known about how their internal processes work from a forensic perspective.
The paper, titled "Foundations for Investigations of Agentic AI Systems from the Forensic Analysis of OpenClaw," takes the OpenClaw system as a case study. It aims to build a foundation for investigating agentic AI by examining this specific system, offering a starting point for understanding how such systems might leave traces or be analyzed after an incident.
Because this is a single preprint, the findings are preliminary. Still, the work addresses a real gap: as agentic AI becomes more widespread, investigators will need reliable methods to examine these systems. The paper's approach could help shape future forensic tools and practices for a new class of digital actors.