AI news ·
DSCI releases India-focused liability guide for agentic AI systems
India's DSCI released a liability guide for agentic AI, mapping responsibility across five enterprise groups and model providers when autonomous systems cause harm.
The Data Security Council of India (DSCI) released an India-focused liability guide for agentic AI systems, mapping how responsibility is assigned when autonomous AI agents malfunction or deviate from their duties. The framework arrives as enterprises across education, government, healthcare, insurance, and legal sectors increasingly deploy AI agents that act independently, raising urgent questions about who bears legal and financial risk when those systems cause harm.
Vinayak Godse, CEO of DSCI, said organisations require "clear decision spaces and instruments they can implement." He added that liability conversations will not lead to practical, sector-specific action unless these factors are integrated across the value chain.
How the framework maps responsibility
The guide applies established binding doctrines to autonomous, multi-vendor AI systems. It traces where responsibility sits across the entire AI use value chain, using concepts of hosting and control over AI models to allocate liability. Architectural control and ownership over decision spaces eliminate external liability shields, the DSCI guidance states. Enterprises would likely become the primary legal owner of both infrastructure and agent behaviour.
Responsibility falls to 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. Model providers form an additional sixth category. Each carries distinct liability questions depending on how the AI enters an enterprise.
Rejecting binary liability models
The report warns against framing liability as an either/or choice between end-user organisations and upstream players. DSCI describes this approach as "legally flawed, technically inaccurate, commercially paralysing, and eroding the level playing field." AI enters enterprises through various routes, and each path demands careful examination rather than blanket assumptions about who pays when something goes wrong.
For professionals in heavily regulated fields, this nuance matters. A hospital deploying an AI scheduling agent, a government office using automated benefits determination, or a law firm relying on document review agents each face different liability profiles depending on their contractual relationships, hosting arrangements, and degree of control over the model's decision-making.
Why this matters for legal, government, and healthcare professionals
The DSCI guide signals that India's regulatory thinking is moving toward shared, context-specific liability rather than defaulting to end-user responsibility. For legal professionals drafting AI procurement contracts, this means liability clauses need to reflect the actual control structure, not boilerplate indemnification language. Government agencies procuring AI systems should audit where decision-making authority sits before deployment. Healthcare administrators and insurers must understand that hosting an AI model internally versus accessing it through a third-party API creates fundamentally different legal exposures. Professionals building expertise in these frameworks can benefit from structured learning such as AI Regulatory Compliance Courses and AI Public Policy Courses that address sector-specific liability analysis.