Who Will Protect Banking Consumers’ Rights in the Age of AI?
By Steve Cocheo, Senior Executive Editor at The Financial Brand
Simple Subscribe
Subscribe Now!
If a consumer contemplated a financial strategy completely out of line with their circumstances, a friend who told them they were a genius wouldn’t be much of a friend. A banker who egged them on even when they realized the consumer was heading for a cliff would violate fiduciary duties.
Key insight: Yet one of the fears about some forms of financial AI is “sycophancy” — AI telling the user what they would like to hear, whether it’s factual or safe, rather than what they should hear.
“A key characteristic of these models that no one has solved for yet is engagement: The more you engage with them, the more they may potentially take you down the path that you want to go as a consumer,” says Delicia Hand, senior director, digital marketplace, at Consumer Reports.
Why this matters: There’s a risk that AI may not confront the consumer with “uncomfortable truths,” possibly leading to harmful consequences.
For this reason, Consumer Reports considers AI sycophancy not a technical glitch, but a form of “honesty failure.”
The issue of honesty and non-manipulation of a consumer by AI is just one facet of risk that Consumer Reports identifies in a massive recent research paper, “AI in Consumer Finance: A Landscape Analysis of Market Forces, Evaluation Frameworks, and Regulatory Gaps.” The report, issued in April, has been followed up by the nonprofit organization’s Consumer Finance AI Standard, a working framework “defining the rights, protections and design practices consumers are owed by AI-powered financial products.”
Why this matters: Warns Hand: “The last thing any of us wants to see is another crisis — one that could have been avoided — because we were over-enthusiastic in deploying models that we didn’t fully understand.”
Need to Know:
- 57% of people surveyed by Consumer Reports in fall 2025 said the current legal structure doesn’t adequately protect them from risks connected to AI in financial services.
- 92% of the sample said that they want to know when AI is involved in financial decision-making.
- 77% want to be able to opt out of AI-driven decisions on major financial issues.
- Nearly six in ten think AI systems should be monitored continually for fairness and accuracy.
- Less than 10% of the sample completely trust financial companies to use AI responsibly.
Reality check: “The introduction of AI into financial decision-making does not create exemptions from civil rights, fair lending or unfair and deceptive practices laws,” argues Consumer Reports.
Consumer Reports makes the point that financial institutions often don’t have a clear idea what is going wrong inside the “black box” — from evolving towards discriminatory lending patterns to giving bad or misleading advice to suggesting options benefiting the provider more than the consumer.
The standards, issued in late June, cover the issues that can arise in detail, from consumer-facing AI to AI deep in the background of financial providers’ operations.
Key insight: “We are in a moment where we’re no longer building technology that we can fully control,” says Hand, an attorney, who spent close to a decade at the Consumer Financial Protection Bureau before joining Consumer Reports. “So, it is worth thinking differently about what we need to do to actually make sure that financial AI is a win-win for all of us.”
Hand says an important aspect of the consumer-facing side of AI is the matter of redress — how the consumer obtains relief if and when the technology errs.
She points out that banks and other providers already have chatbots, voice assistants and other tech aids that can plunge a consumer into a “doom loop.”
The new doom loop: “So what does that look like when there are very intelligent models that consumers deal with?” she asks. “Is there a meaningful escalation of problems when you are on the phone with a much-more-sophisticated algorithm that has the ability to make decisions about how and where to direct your issue?”
Hand is concerned that AI will sometimes decide that it can handle the consumer’s complaint and will decline to send the consumer to a human “off-ramp.”
On the other end of the process, where agentic commerce and payments begin to become available routinely, Hand worries about disclosures. Will an agent take any note of a disclosure? And does it matter, if the consumer never sees it because an agent has taken charge? (Some bankers and consumers will point out that in spite of best intentions, extensive disclosures required for human-to-human financial dealings may be there, but often don’t get read.)
Read more: Why Bank App Virtual Assistants Still Need Assistance Themselves
A Nine-Point Plan for Financial AI Consumer Protection
Hand says Consumer Reports’ standards were put together in consultation with multiple stakeholders and “road tested” with former financial services regulators and some members of the broad financial services community. She adds that the group is looking for additional input and regards the document as a work in progress.
Why the standards are significant: “We’re not necessarily looking for the standards to be taken up by regulators per se, but we’re looking to influence the conversation and to get folks thinking about these issues sooner rather than later,” says Hand. “That’s because this marketplace and this technology are moving so quickly.”
Current state and federal policies, laws and regulation on AI in general are a patchwork with many gaping holes.
These are the nine points of Consumer Reports’ lengthy standards document, summarized:
- Security and trust. AI products must minimize risk and protect from misuse, data breaches and unauthorized financial harms.
- Privacy and data minimization. Data is collected, used and shared responsibly and only to the degree necessary to deliver service and to maintain compliance. Consumers retain control over data without friction.
- Transparency and accountability. Providers of AI financial products and services communicate with consumers.
- Honesty and non-manipulation. Beyond the issue of sycophancy, AI financial products must be honest with users and independent of both their own leanings and providers’ own interests.
- Reliability and operational integrity. AI financial products and services should produce consistent, accurate and predictable outputs.
- Consumer agency and control. Users should be able to understand, question and influence AI decisions that affect them. This includes being able to override decisions made on their behalf by AI systems such as agents.
- Duty of loyalty. The primary purpose of AI products and services should fulfilling the consumer’s interests, not the provider, its partners nor third parties.
- Fairness and nondiscrimination. AI should not produce direct or indirect discriminatory results.
- Duty of vigor. From the body of the report: “The AI actively fights for you — spotting the rights you have and arming you to use them, so you don’t have to know the law yourself.”
Perspective on data usage: These standards come even as federal regulation of open banking data sharing remains under revision.
“Our view is that providers show the full value before the full ask [for data],” says Hand, “and then they should be measured in the ask for more data. There’s already an asymmetric relationship between a consumer and any financial institution. So we advocate that it should be a smaller and a fair exchange.”
Read more: Retail Bankers Are Adopting AI for All the Wrong Reasons
What Should Banks and Credit Unions Do with the Standards?
Major banks, credit unions and fintechs have compliance staffs to evaluate and work with the standards, which at this point are only something to consider and not requirements.
But many smaller players don’t have that depth of expertise on staff.
What should they do with this guidance? Hand suggests they can use the standards as an aid for vendor relations, both with existing relationships and ones under consideration.
“The standards are a yardstick to hold vendors to without having to develop the expertise in-house,” says Hand. She believes they can give institutions some leverage when discussing AI with vendors.
Questions can be prompted by the standards. Some examples she suggested, in the case of an AI financial assistant under evaluation:
• Does the assistant stay honest under pressure?
• Is data use minimized? And is it contained and not repurposed for training the AI model?
• Are there issues that aren’t apparent that the customer may want to know more about?
From Hand’s perspective, a key priority is honesty and non-manipulation, the fourth standard in the list above.
Read next: Your Customers Will Blame You When Their Shopping Bots Go Rogue
