AI profiling in finance needs stronger consumer safeguards: The growing debate over technology addiction has brought renewed attention to how digital platforms influence behaviour. In August 2026, Meta agreed to settlements worth up to $18 billion in the United States over allegations that Facebook and Instagram were designed to encourage addictive use among children. Meta denied wrongdoing. In India, the Delhi High Court asked the Centre in September whether it was considering a policy to regulate potentially addictive design features such as autoplay and infinite scrolling.
These developments raise a larger concern about commercially optimised algorithms operating at enormous scale, often before legal safeguards have caught up. With artificial intelligence increasingly integrated into everyday services, the consequential question is what happens when digital systems move from predicting what users will click to inferring their personalities, vulnerabilities and financial preferences, and deciding how to engage with them.
READ | Women’s jobs face greater disruption from artificial intelligence
Financial services provide a particularly important setting for examining this concern. AI is increasingly used in credit underwriting, fraud detection, customer service and risk assessment. Its outputs can influence whether a person receives credit, how a transaction is treated or how a financial institution assesses risk.
AI is also reaching consumers directly. Individuals increasingly turn to general-purpose chatbots for financial guidance, often without the protections associated with regulated financial advice.
A September 2026 study by Saturn, a UK technology provider to financial advisers, tested 18 AI models using more than 10,000 responses to personal finance questions. It found that answers were incorrect or incomplete 57% of the time, with the failure rate rising to 88% for complex questions. The findings come from one industry-sponsored study, but they illustrate the risks of relying on AI outputs for consequential financial decisions.
The Reserve Bank of India’s FREE-AI Committee report, released in August 2025, recognised the risks associated with AI in high-stakes financial applications. Its recommendations seek to encourage innovation while protecting consumers and maintaining public trust.
However, a framework centred on responsible deployment cannot, by itself, settle questions about the rights of individuals subjected to AI-driven profiling. What safeguards should govern systems that infer personality traits, risk preferences or emotional states? When should a financial institution be allowed to use such inferences to influence customer decisions? And how should commercial incentives be regulated when conversational AI also serves as a channel for advertising and product promotion?
How AI profiling can influence financial decisions
The concern is grounded in what is known about human decision-making. Behavioural economics has established that financial choices are influenced by emotions, cognitive biases and individual characteristics. People do not always assess risk or process financial information in the manner assumed by conventional models of rational choice.
Research examining personality and investment decisions has found associations between certain traits and the tendency to follow others’ investment choices, commonly described as herd behaviour. Such findings acquire a different significance when combined with AI-driven profiling.
If personality traits and behavioural tendencies influence financial decisions, systems capable of identifying these characteristics at scale could be used to predict how consumers will respond to particular products, messages or incentives. The same capability could also be used to influence their choices.
Personalisation has already become an important part of financial services. Banks and fintech companies use customer data, machine learning and predictive analytics to tailor products, improve service and identify potential customer needs.
Recent corporate disclosures illustrate the extent of this transformation. Bajaj Finserv’s 2025-26 annual report describes its FINAI strategy, including the use of digital and voice signals to generate behavioural insights and develop models for credit risk, personalisation and propensity scoring.
These applications can make financial products more relevant and accessible. Better risk assessment can also help lenders serve customers who might otherwise be excluded from formal credit.
Yet the benefits of personalisation do not remove the need for limits on its use. Privacy rules govern the collection and processing of personal information. They are less explicit about the legitimacy of inferences drawn from that information and their use in shaping financial choices.
This distinction becomes important when customer profiling moves beyond identifying preferences to assessing behavioural vulnerabilities.
READ | Can artificial intelligence strengthen democracy? Italy offers a test case
From personalisation to deep behavioural profiling
Early recommendation systems relied largely on predefined rules. Machine learning made it possible to identify patterns across much larger datasets and adapt recommendations as new information became available.
The emerging frontier is emotional and behavioural profiling. Systems can attempt to infer a person’s mood, impulsivity, susceptibility to particular incentives or willingness to take risks. These inferences may then inform the timing of notifications, the design of an interface or the financial products presented to the customer.
The distinction between useful and harmful personalisation depends partly on context. Recommending music based on an inferred mood generally carries limited consequences. Using similar techniques to influence trading behaviour, borrowing decisions or the purchase of complex financial products raises more serious concerns.
An engagement system optimised for commercial returns may identify that certain users respond more strongly to repeated prompts, rewards or emotionally framed messages. If these responses increase the outcome being optimised, the system may reinforce them without being explicitly programmed to exploit a vulnerability.
Researchers have described related risks as inadvertent algorithmic exploitation. The problem arises from the interaction between the system’s objective and the behavioural patterns it identifies. Harmful outcomes do not necessarily require an intention to manipulate.
In financial services, the consequences can extend beyond excessive time spent on a platform. Behavioural profiling could influence how frequently individuals trade, how much they borrow or whether they accept unsuitable financial products.
When gaming behaviour becomes a financial risk score
A patent held by Wells Fargo illustrates how far behavioural inference can extend. The US patent, titled Utilising Gaming Behavior to Evaluate Player Traits, was granted in July 2019. It describes a system that analyses an individual’s computer-gaming behaviour to infer personal characteristics and calculate a financial risk factor.
The information contemplated includes the types of games played, the frequency and duration of play, decisions made during gameplay and conduct such as using cheat codes or modified versions of games.
These observations can be used to infer characteristics such as risk propensity, goal orientation and other behavioural traits. The patent describes how the resulting assessments could inform decisions concerning loans, credit facilities and investment services.
The existence of a patent does not establish that the technology has been commercially deployed. Nevertheless, the patented system demonstrates how information generated in one context can be converted into an assessment with consequences in an entirely different part of a person’s life.
The underlying regulatory problem is substantial. A person may consent to data being collected for one service without anticipating that the same information could contribute to judgments about their creditworthiness or suitability for financial products.
Such inferences may also be unreliable. Behaviour within a computer game does not necessarily establish how a person will manage a mortgage or repay a loan. Yet an automated system could assign economic significance to these observations, particularly if its assessments are difficult to examine or challenge.
Regulation must therefore extend beyond identifying what personal information was collected. It must also consider what was inferred, whether the inference was justified and how it affected the individual.
READ | Can India power the artificial intelligence dream?
AI regulation must address inferred characteristics
India’s Digital Personal Data Protection Act, 2023 provides an important legal foundation for the processing of digital personal data. Its provisions on lawful processing, consent and the rights of individuals are relevant to AI systems that use personal information.
However, the Act does not establish a dedicated framework of rights and safeguards for algorithmic profiling and consequential automated decisions. Compliance with data protection requirements does not necessarily resolve whether a behavioural inference is accurate, appropriate or fair.
This leaves a regulatory question that cannot be answered through voluntary responsible-AI principles alone.
India need not adopt a sweeping AI statute to begin addressing these risks. Existing financial regulators, including the RBI and Securities and Exchange Board of India, can develop more specific safeguards for high-risk applications within their respective jurisdictions.
One priority should be identifying forms of behavioural categorisation and profiling that require restrictions, particularly where inferred vulnerabilities may be used to influence borrowing or investment decisions.
A second is stronger purpose limitation. Data and models developed for legitimate functions such as fraud detection or credit assessment should not be repurposed for marketing or engagement optimisation without appropriate legal justification and safeguards.
Regulators should also require high-risk AI systems to be assessed for their capacity to exploit behavioural vulnerabilities. Conventional assessments of model accuracy, bias and technical performance may miss harms arising from how a system interacts with consumers.
For consequential financial decisions, individuals should have meaningful avenues to obtain explanations, contest inaccurate assessments and seek human review. The requirements can be proportionate to the risk involved, with stricter oversight where profiling affects access to credit, investment decisions or financial well-being.
The RBI’s FREE-AI framework provides a starting point. Its effectiveness will ultimately depend on whether broad principles translate into enforceable obligations for financial institutions and technology providers.
As AI becomes capable of drawing increasingly detailed conclusions about consumers, the question of who controls those inferences, and for what purposes they may be used, deserves the same regulatory attention as the personal data from which they are derived.
Deepti George is the founder of Yutadhi, a think tank focused on financial services, and Aashna Kothiyal a lawyer and policy advisor.