Mumbai : As AI tools become more common, concerns are growing about how they are used and the risks they may create. This is especially important for banks and financial institutions that handle sensitive customer and financial data. To address these concerns, the Reserve Bank of India (RBI) has proposed new guidelines that would require all banks to have a “kill switch” for AI systems. This means banks must be able to quickly turn off any AI model if it starts giving wrong, harmful, or unsafe results.
The RBI released draft guidelines on Wednesday that set out rules for banks and other regulated entities using AI. Under the proposal, banks must have systems in place to override, suspend, or completely deactivate AI models whenever needed. The central bank said no AI system should operate without the ability to be shut down immediately if it produces harmful or incorrect outputs. The move comes amid growing concerns about AI-related cybersecurity risks, especially after issues linked to Anthropic’s Claude Mythos AI model.
The RBI also said that AI-based decisions must always be monitored by humans. Even if AI is performing a task, a person must oversee important decisions. The proposed rules would apply to all types of models used by banks, from simple spreadsheet-based tools to advanced AI systems. Banks will be fully responsible for the results produced by any model they use, whether the model was developed internally or bought from a third party. They must also carefully evaluate any AI model before using it.
The RBI wants banks to classify their AI models according to risk levels and apply appropriate checks and controls. If a model poses risks beyond what the bank is willing to accept, the bank must take immediate action. This could include adding more controls, limiting its use, fixing problems, or shutting down the model completely.
Banks must review the risk level of each model at least once every year. High-risk AI models will need approval from the Board’s Risk Management Committee before they can be deployed. For the first time, the RBI has made AI governance a responsibility of a bank’s board of directors. Every regulated entity must have a board-approved Model Risk Management Framework covering all AI and other models used in the organization.
The RBI said that as banks increasingly use AI and machine learning in business operations and decision-making, weak controls could expose them to financial, operational, legal, and reputational risks. The central bank warned that poorly managed AI systems could lead to wrong decisions, financial losses, operational problems, regulatory violations, and harm to customers and the wider financial system.
The RBI also highlighted the risk of relying too heavily on a small number of AI providers and said banks must manage such supply-chain risks carefully. They must also ensure that AI systems do not create new security vulnerabilities. For customer-facing AI systems, banks must clearly inform customers when they are interacting with AI and must give them the option to speak with a human at any time.
The draft guidelines also warn against “automation bias”—a situation where employees trust AI outputs too much without using their own judgement. For generative AI systems that interact with customers, the RBI has called for stronger cybersecurity measures. The RBI has invited public feedback on the draft guidelines until July 24.
Read Also : PM Modi Marks 51st Emergency Anniversary, Condemns 1975 ‘Assault on Constitution’