Announced 30 Oct 2026 · Sources checked
What the guide addresses
The Data Security Council of India (DSCI), an industry body focused on data protection and cybersecurity, released a guide on 30 October 2026 addressing liability for agentic AI systems. The core question: who is responsible when an AI agent that makes autonomous decisions on behalf of a user goes rogue, deviates from its instructions, or causes harm?
The guide defines accountability frameworks based on the level of agent autonomy. A fully autonomous agent that makes decisions without human approval creates a different liability profile than a semi-autonomous agent that checks with a human before executing high-risk actions. The guide maps these profiles to existing Indian legal frameworks including contract law, tort law and consumer protection statutes.
Why now
Agentic AI — AI systems that can execute multi-step tasks autonomously, such as booking travel, managing finances, filing legal documents, or controlling infrastructure — is moving from research demos to commercial deployment. As agents become capable of causing real-world harm (financial loss, privacy violations, physical damage), the liability question becomes urgent.
India is one of the first countries to produce sector-specific guidance on agentic AI liability. The guide follows the TRAI telecom AI accountability draft (also released this week) and positions India as an early mover in AI governance alongside its AI for All strategy and the IndiaAI mission.
What it means for enterprises
For enterprises deploying AI agents in India, the guide defines clear expectations: organizations need to map the decision spaces of their agents, define the level of human oversight for each agent function, document the allocation of responsibility between developer, deployer and user, and maintain audit trails of agent decisions.
The guide does not have the force of law, but DSCI recommendations frequently inform Indian regulatory frameworks. Companies that align their agent deployment practices with the guide now will be better positioned when formal regulation arrives, which the TRAI draft and the government's broader AI strategy suggests is coming.
Caveats
The guide is an industry body document, not government regulation. It provides a framework for enterprises to assess liability risks but does not create new legal obligations. The specific application of Indian law to agentic AI scenarios remains untested in court.
AI Lookout covered the reporting on the guide; it did not obtain the full DSCI document.
The autonomy spectrum
The DSCI guide uses a spectrum of agent autonomy to map liability. At one end are simple tools that execute predefined actions with no discretion — the developer is clearly liable if the tool malfunctions. At the other end are fully autonomous agents that make independent decisions using LLMs, APIs and tool access — here the liability question is more complex and may depend on the degree of human oversight, the specificity of the instructions and the foreseeability of the agent's actions.
The guide proposes that enterprises classify their agent deployments along this spectrum and implement governance measures proportionate to the autonomy level. A customer service chatbot that only retrieves pre-approved answers presents lower liability risk than an agent that can independently negotiate contracts or execute financial transactions, even if both use the same underlying AI model.
International context
India's focus on agentic AI liability follows similar moves in other jurisdictions. The EU AI Act includes provisions on liability for AI systems, though it does not specifically address agentic autonomy. The UK Law Commission has consulted on AI liability reform. Japan and Singapore have published AI governance frameworks.
What distinguishes India's DSCI guide is its explicit focus on agents — systems that act autonomously on behalf of users — rather than AI systems in general. As agents become more common (booking travel, filing taxes, managing subscriptions), the need for clear liability rules becomes more acute. The DSCI guide is one of the first documents anywhere to address this directly.
What enterprises should do
The DSCI guide recommends several practical steps for enterprises deploying agentic AI: classify each agent by autonomy level and potential harm, define the human oversight required for each level, document instructions and decision boundaries for the agent, implement logging and audit trails for agent decisions, and regularly test agents against their instructions.
The guide also recommends that enterprises establish clear contractual terms with AI vendors specifying liability allocation for agent failures. If an AI agent provided by a vendor makes an unauthorized transaction, the contract should state whether the vendor, the deploying enterprise or the end user bears the loss.
What this means for the global AI liability conversation
India's DSCI guide joins a growing body of international guidance on AI liability that includes the EU's Product Liability Directive amendments, the UK's AI Liability consultation and the US NIST AI Risk Management Framework. Each takes a different approach — the EU focuses on strict liability for high-risk AI, the UK on existing tort law application, the US on voluntary risk management and India on sector-specific agentic liability.
For global companies deploying AI agents, the divergence in liability frameworks across jurisdictions creates compliance complexity. An AI agent that meets India's DSCI guidance may not automatically satisfy the EU's product liability regime or US tort standards. The DSCI guide is a starting point for enterprises to map their agent deployments against emerging international standards.
Common questions
What is the DSCI agentic AI liability guide?
An industry framework defining who is responsible when an autonomous AI agent goes rogue or causes harm.
Does it have legal force?
No — it is industry guidance, not regulation, but DSCI recommendations often inform Indian policy.
Who is liable for an agent's actions?
It depends on autonomy level — the guide defines frameworks for developer, deployer and user liability.
Sources & further reading
How this story was made
Written by Kristian Kostov with AI assistance and checked against the linked sources. Company performance claims are attributed to the company. Analysis reflects AiLookout’s interpretation; we have not independently tested the products discussed. The cover is an AI-generated editorial illustration, not a photograph or evidence of an actual event.
Our editorial standards
