Your AI Agent Made the Mistake, Who Is Accountable for the Damage?
ET Edge Insights, Monday, November 2nd, 2026
DSCI guidance says liability for AI agent mistakes depends on who controlled the agent's permissions, architecture, deployment and actions.
As AI moves from answering questions to executing decisions, India's enterprise AI boom faces an accountability test.
A Data Security Council of India guidance paper argues liability will depend not only on who built the model but on who controlled the agent's permissions, architecture, deployment and actions.
Using the example of a customer service agent that issues a refund beyond its authorized limit, the article asks whether fault lies with the model, a vendor's unsafe defaults, an integrator's excessive permissions or an enterprise that lacked safeguards.
Agentic AI makes failures harder to trace because agents plan tasks, choose tools and act across multiple systems.