Introducing our new whitepaper, Consumer Welfare in the Age of Generative AI: From Theory to Practice.
We asked more than 4,000 Americans a straightforward question: do current laws protect you from the risks of AI in financial services? A majority – 57 percent – said no. Fewer than one in ten completely trust financial companies to use AI responsibly. And when we asked people to weigh the benefits of AI in finance against the risks, more said the risks come out ahead than said the benefits do; the single largest group just wasn’t sure.
Perhaps, that’s not a public afraid of technology, it’s a public reading the situation accurately. People can feel that something has changed in how decisions about their money get made, and they can feel that the rules haven’t caught up. Our new whitepaper is about that gap – what’s driving it, what it costs consumers, and what it will take to close it.
Why the old model doesn’t fit
For most of its history, consumer protection has worked like product safety. Find the defective product, warn people, compel a fix or a recall. That approach assumes three things: that harm traces back to a specific product, that a consumer can see the risk and steer around it, and that the product holds still long enough to be evaluated.
AI in financial services breaks all three. These systems aren’t static products – they’re adaptive, and they change their behavior as they take in new data. The harm doesn’t sit in one faulty part; it’s embedded in how a system prices a loan, scores an applicant, or routes a complaint. And most people can’t tell when AI is being used on them at all, let alone judge whether it’s working in their interest.
AI is no longer a feature bolted onto a financial product. It is becoming the infrastructure of the market itself – shaping which consumers see which products, at what price, on what terms, and with what quality of service. Once that’s true, “is this product safe?” is still a necessary question, but it stops being a sufficient one. The question has to become: does this system treat people fairly, transparently, and accountably?
What the paper does
The whitepaper moves the conversation from cataloguing risks to defining obligations. It does the following things:
It documents the gap, drawing on our 2025 nationally representative AI in Financial Services Survey, our landscape analysis of existing frameworks and regulation, and a decade of CR testing and advocacy in digital finance. It introduces a Consumer Welfare Standard – a practical, obligation-based framework that translates established consumer-protection principles into what AI systems actually owe the people they serve, organized around five categories:
- Information Integrity – what the system tells you is accurate, calibrated, and not engineered to manipulate.
- Fair Treatment – you’re judged on your real financial situation, not discriminated against and not exploited.
- Consumer Control – you stay in charge, especially as AI begins to act on your behalf.
- Accountability and Remedy – when the system gets something wrong, someone answers for it and you can get it fixed.
- Systemic Responsibility – the broader system, from the AI supply chain to resilience to environmental cost, doesn’t quietly push its costs onto the public.
And it maps the policy landscape, identifying concrete openings for policymakers, regulators, and industry to close the most serious gaps.
The part that should worry us most
The hardest problems are still ahead. AI is no longer just deciding – it’s beginning to act, completing transactions on a consumer’s behalf. Existing consumer protection law largely assumes a person affirmatively initiates a transaction. When an agent can initiate, execute, and finish one without a human in the loop each time, the foundational ideas of consent, authorization, and liability need to be rethought. The paper treats agentic AI as the next frontier of consumer risk for exactly this reason, and argues the time to set the rules is before these systems are everywhere, not after.
The legal backdrop makes this more urgent, not less. Statutes like ECOA, FCRA, and the prohibitions on unfair and deceptive practices apply to AI in full – but they were written for a slower, more legible market, and they’re being applied unevenly. States like Colorado, California, and New York are moving to fill the gaps, which is welcome, but the result so far is fragmentation rather than coherent protection. At the same time, some federal protections are narrowing. The outcome-based fairness test that anchored fair lending for half a century is being pulled out of federal ECOA enforcement this summer. When the public floor drops as the technology accelerates, an independent, consumer-first standard matters more.
What we’re asking for
To be clear about what this paper is not: it is not a call to ban AI in financial services. It is a call to hold AI to the same obligation any provider of consumer financial services already carries – to serve the consumer’s interest honestly, fairly, and accountably. Meeting that obligation takes work on four fronts at once: policymakers updating the legal floor, regulators building the technical capacity to actually supervise these systems, industry moving from aspiration to pre-deployment testing and transparent reporting, and independent evaluators building the infrastructure that makes AI’s impact on consumers visible and comparable.
This whitepaper is the case. Our Consumer Finance AI Standard is the operational answer to it – what good looks like at the level of a specific product. Read together, they’re an invitation: to every company, regulator, funder, and advocate who believes AI-mediated finance should come out in consumers’ favor, this is the work, and we’d like to do it with you.
