Are AI finance apps already nudging users toward risky borrowing?

Within a broader discussion of consumer-autonomy risk, the IMF note by Sonja Davidovic and Hervé Tourpe cites research (Aldasoro and others 2024) showing that harm from AI-driven financial nudging is not hypothetical: "these dynamics have already arisen in the context of some AI-driven personal finance apps, which have been criticized for recommending cash advance offers and promoting borrowing products when a user's account balance dips below particular thresholds, even in situations where those products would worsen consumers' long-term financial outcomes." The note treats this as the concrete, present-tense instance of a risk category it otherwise describes prospectively.

The note's broader reasoning is that ongoing data access plus autonomous action compounds the scale of harm: "ongoing access to financial data, ability to engage in behavioral nudging, and capacity for autonomous action can potentially increase the scale, subtlety, and personalization of activities that run counter to consumers' interest." It places the cash-advance example alongside a wider list of ways agentic systems can fail consumers — they "may misinterpret user intent, optimize provider incentives rather than user welfare, drift from original objectives over time, or engage in subtle behavioral nudging at scale" — and alongside systemic risks like algorithmic herding and prompt injection, but the personal-finance-app example is the one the note anchors to an existing, criticized practice rather than a projected one.

This supplies a concrete mechanism behind the caution Do consumers want to approve AI shopping purchases before they happen? finds at the level of stated preference: RTB House reports that baby boomers in particular want to sign off before an agent spends money, and the IMF's cited example of apps pushing cash advances when a balance runs low is a documented case of exactly the kind of provider-aligned, consumer-unfriendly nudge that preference is guarding against. It also sharpens the stakes in Should AI agents make payment decisions independently?: the case for separating probabilistic agent judgment from deterministic execution is not just about preventing errors, but about preventing a financially interested system from exploiting a vulnerable moment before a human reviews the action.

The excerpt gives no detail on the cited study — no sample, no app names, no measure of how often this nudging occurred or its scale — only the citation "(Aldasoro and others 2024)." It also does not say whether the apps involved met the note's own definition of agentic AI (autonomous, multistep, delegated) or were simpler rule-based personalization engines, which matters for the note's broader argument that agentic autonomy specifically increases this risk. What the excerpt supports is narrower: nudging toward products that worsen outcomes at moments of financial vulnerability has already been observed in AI-driven finance apps, which the IMF treats as evidence that agentic autonomy would amplify an existing problem rather than invent a new one.