Skip to main content

RBI flags AI-agent risks to bank systems

26 OCTOBER 2026·2 MIN READ·3 SOURCES

The Reserve Bank of India has alerted lenders and the central government to risks associated with AI agents in banking. Governor Sanjay Malhotra said the regulator and commercial lenders are working to identify and fix IT vulnerabilities.

RBI flags AI-agent risks to bank systems

Key takeaways · 4

  • 01

    Review agent permissions and software patching: the report identifies excessive administrative rights and unpatched flaws as possible routes to sensitive financial data.

  • 02

    Treat agent activity as distinct from ordinary chatbot use: agents can independently carry out complex, multi-step actions within predefined parameters.

  • 03

    Keep reported incidents separate: the SBI records exposure involved 3,897 customers, but available evidence does not directly link it to AI agents.

  • 04

    Track RBI’s resilience work alongside its warning: the Bhubaneswar institute is intended to train financial-sector staff and research infrastructure security, robustness and efficiency.

RBI warns about agent risks

The Reserve Bank of India alerted lenders and the central government to risks associated with AI agents, which the report describes as a new cybersecurity challenge for banking.[1] Governor Sanjay Malhotra said the regulator and commercial lenders are working to identify and fix vulnerabilities in IT infrastructure, and stressed the need for stronger safeguards against AI-driven cyberattacks.[1] The concerns include unauthorized access to customer records and disruptions to essential banking operations.[1]

Why agents raise different concerns

The article distinguishes AI agents from traditional chatbots by their ability to execute complex, multi-step actions independently, using predefined parameters.[1] Reported activity has raised concerns that automated programs could exceed users’ instructions or initiate unapproved transactions.[1] The article also cautions that not every automated agent is inherently threatening.[1]

Reported incidents and limits

The report says overseas test environments produced accounts of AI agents targeting banking systems and extracting data, while major U.S. banks reportedly detected third-party AI agents operating within customer accounts.[1] It says technical details of the overseas test breaches had not been disclosed.[1] Separately, State Bank of India acknowledged a suspected breach involving records of 3,897 customers posted to the dark web.[1] The exposed files contained personally identifiable information that criminals could use for identity theft, phishing and unauthorized transactions, but the available evidence did not directly link the leak to AI agents.[1]

Bhubaneswar resilience plans

In related infrastructure plans, Malhotra said a Bhubaneswar institute would train RBI staff and employees of banks and other financial-sector organisations, and conduct research to improve the security, robustness and efficiency of financial infrastructure.[3] The new data centre will support RBI applications including NEFT and RTGS, the e-Kuber government-banking platform and services for regulated entities.[3] Malhotra said two-thirds of the complex’s land was allocated to the institute and one-third to the data centre.[3] He said Bhubaneswar was selected partly because it is in a different seismic zone from RBI’s three existing data centres in Maharashtra.[3]

Banks and their technology partners have reason to examine what permissions automated agents receive and whether software flaws could expose sensitive data. The report also makes an important distinction for risk assessments: it describes AI-related concerns and a separate SBI records leak, but does not establish a link between them.

Why it matters
Daily session

Bring this to your team — one session a day, built for your industry.

kju.ai is built for teams: daily sessions, team analytics, and rollout support for your organization.

For teams

How this developed

  1. 26 October 2026

    RBI flags AI-agent risks to bank systems

Sources

AI fluency, one session a day, built for your work.