By Gopika Gopakumar
MUMBAI, June 24 (Reuters) – India’s central financial institution has proposed guidelines requiring banks to strengthen oversight of dangers tied to AI and machine-learning fashions, mandating board-approved insurance policies, stronger controls and mannequin inventories.
The Reserve Financial institution of India stated banks should put in place a board-approved threat administration framework protecting all fashions, together with these for AI and machine-learning.
Regulated entities should assess threat at each the person mannequin degree and throughout the enterprise on an ongoing foundation, the RBI stated.
If dangers are discovered to be extreme, lenders ought to take well timed corrective steps, together with enhanced controls, restrictions on use, remediation or decommissioning of the mannequin, and submit a report back to the board’s threat administration committee, the financial institution added.
The RBI additionally stated banks ought to guarantee all fashions, together with third-party fashions, are topic to impartial validation.
Banks ought to set up human oversight for AI fashions utilized in automated decision-making, the draft pointers stated. For generative AI fashions that interface with clients or exterior customers, extra cybersecurity controls needs to be applied.
RBI has invited suggestions on the draft pointers by July 24.
(Reporting by Gopika Gopakumar; Enhancing by Shailesh Kuber)
