Originals · Startup · Startups & Fintech
DSCI maps who is liable when agentic AI systems go wrong
The industry body says India does not need brand new AI liability law first; it needs clear allocation across deployers, integrators, product firms, tool makers and cloud providers.
On 30 October 2026 the Data Security Council of India released an India focused AI Liability Industry Guidance for Agentic AI. Business Standard and Express Computer both report that the paper asks who bears the consequences when autonomous AI agents cause errors, operational disruption or other harm, and frames the document as a live discussion paper rather than a binding regulation.
DSCI's press note and PDF say the guide does not invent a new regulatory regime. It maps established doctrines such as duty of care, causation, apparent authority and product liability onto multi vendor agentic stacks. Responsibility is traced across five groups that bring AI into use: user organisations, service providers and integrators, product companies and startups, tool and framework providers, and cloud and infrastructure providers, with model providers treated in a separate chapter.
CEO Vinayak Godse said organisations need decision spaces and instruments they can actually implement, or liability talk will not become sector practice. The guidance warns against treating liability as a binary between end users and upstream vendors, and it discusses graded controls under a prevent first, halt last hierarchy rather than a single kill switch. Industry leaders are invited to test the guide against real deployments and send feedback so DSCI can revise it.
For bharat.digital readers the document sits next to CERT-In warnings on AI assisted payment API abuse and RBI cyber rules for non bank payment operators. It gives CISOs and product counsel a shared Indian reference for agentic AI contracts while MeitY and sector regulators keep writing their own rules. It is guidance, not law, so courts and contracts still decide each case.
