Authors: Associate Professor Dr. Pramod Kumar, Assistant Professor Mr. Mohit Shrivastava, Assistant Professor Mr. Saood Iqbal khan, Assistant Professor Mr. Lakshya Bhardwaj
Abstract: There are many autonomous artificial intelligence agents which can plan, carry out and adapt multi-step digital activities on their own. This is far outpacing the legal and technical tools that the law enforcement relies on to investigate and prosecute cybercriminals. Self-organizing agents make a series of coordinated decisions, call external agents, and produce emergent behaviour that is non-linear and not totally programmed or even anticipated by the programmer. This poses a challenge to forensically trace actions back to a human agent. The paper will examine the state of forensic readiness, or the organisational and technical ability to gather and present digital evidence for evidentiary purposes in the context of a digital investigation, and question its suitability for an agentic AI world. This international and national legal instrument provides a survey of the Council of Europe's Budapest Convention on Cybercrime, the EU Artificial Intelligence Act, data protection law and comparative evidence rules from the US and India. The purpose is to evaluate the current status of the immunity and liability in the case of an autonomous agent performing the proximate act instead of a natural person. A layered approach offering mandatory action logging, cryptographic provenance mechanisms, clear allocation of liability among developers, deployers and operators, as well as international cooperation mechanisms scaled to machine speed are required for the forensic readiness of autonomous agents. The paper finishes by providing a forensic readiness framework for the agentic AI lifecycle, as well as legislative, regulatory and institutional reforms to fill the upsurge of an accountability gap.
