Hemant Sood, Founder and Managing Director of Findoc Investmart Pvt. Ltd, succinctly summarizes the primary risk: “AI chatbots are excellent for enhancing financial literacy, but the danger arises the moment a fluent answer is mistaken for regulated, personalized advice.”
The quality of the prompt matters
The reliability of an AI-generated financial response largely hinges on how the question is posed. If users fail to give the appropriate context or omit crucial details about their financial situation, the chatbot’s output may be incomplete or misleading, potentially resulting in financial setbacks.
Navy Vijay Ramavat, Managing Director at Indira Securities, notes that intent is rarely the issue: “The intent here is fine. It’s the execution that typically falls short.” He advises a fundamental principle: “AI is designed to assist you, not to do your thinking for you.”
The danger of oversharing personal data
In their quest for more accurate responses, many users unintentionally disclose sensitive personal and banking information to AI tools, including Aadhaar numbers, PAN numbers, credit or debit card details, bank account specifics, income and debt information, phone numbers, and email addresses. Once shared, this information may not always remain confidential, considering how some AI platforms treat user data.
Ramavat cautions that this tendency is riskier than it appears: “Feeding it unmasked financial information is one of the simplest ways to expose yourself to fraud.”
Amit Suri, CFP, Founder and CEO of AUM Wealth, underscores the risks associated with compromised data: “Financial information such as PAN, Aadhaar, bank account details, investment holdings, tax records, or passwords can be exploited for identity theft, financial fraud, unauthorized transactions, phishing attempts, or social engineering scams.”
Arun Patel, Founder & Partner at Arunasset Investment Services, concurs with this concern, highlighting how leaked information can be weaponized: “They exploit real personal/financial details to make the scam appear credible.”
Patel is transparent about his reason for refraining from sharing sensitive information with AI tools entirely: “We cannot ascertain how secure our data and conversations with the platforms are.”
How to use AI responsibly for financial queries
Given the evolving nature of AI tools, which are prone to errors, users must exercise caution when using them for financial topics.
Review the platform’s privacy and data retention policies: Paid or enterprise versions of AI tools usually offer better data protections than free options, so it’s wise to check the level of protection your version provides. Sood warns that this does not guarantee safety: “A paid subscription can’t alone ensure the confidentiality of data shared.”
Switch off data-training permissions: Most AI platforms permit users to decline having their data utilized for model training, although this option is often hidden in privacy settings and may require some effort to find.
Strip out personal details before uploading documents: When providing bank statements, bills, or other financial documents to an AI tool, sensitive data such as names, account numbers, addresses, transaction dates, and merchant names should be omitted. Only share the minimum necessary information. Ramavat suggests a similar caution for text prompts: “Never reference yourself, your family, or acquaintances by name. Use placeholders instead.”
Regarding accuracy, Sood provides an example of AI lagging behind on tax regulations: “We’ve observed chatbots quoting the pre-budget equity LTCG rate of 10% well after it was revised to 12.5% in July 2024.”
Suri agrees that AI-generated responses concerning tax or regulatory issues should not be accepted at face value: “A confident response is not necessarily correct.”
Patel takes this a step further, questioning accountability when AI provides incorrect information: “With AI, who is responsible?”
In conclusion, AI tools are most beneficial when used to support research, analysis, and decision-making rather than acting as a replacement for professional advice and independent judgment. As Sood states: “AI should function at the education level, not the decision level.”
