The Ministry of Electronics and Information Technology (MeitY) in India has proposed new guidelines requiring human-in-the-loop interventions for financial transactions initiated by agentic AI systems. As AI agents become increasingly capable of performing autonomous tasks, including managing bank accounts and executing payments, the government is seeking to establish a safety net to prevent unauthorized or erroneous financial transfers. This move reflects a growing global trend of balancing AI innovation with consumer protection.
Agentic AI refers to software systems that can set their own goals and execute complex sequences of actions without constant human prompting. While this technology promises significant efficiency gains for businesses and individuals, it also introduces risks regarding accountability and security. If an AI agent makes a mistake or is compromised by malicious actors, the lack of a human check could lead to irreversible financial losses.
Under the proposed framework, any AI-driven payment process would require a verified human confirmation before the final transaction is processed. This ensures that users retain ultimate control over their finances, even when delegating routine tasks to automated software. The policy aims to clarify the legal liability of AI developers and service providers in the event of a system failure.
Industry stakeholders are currently reviewing these proposals to understand the technical implications for their platforms. Implementing such a mandate could require significant changes to existing AI architectures, potentially slowing down the deployment of fully autonomous financial assistants. The government is expected to hold further consultations to refine these requirements before they are formally adopted as part of India's broader digital regulatory framework.