The Vanishing Human
Picture this (an illustrative scene, not a real case). It is late. You are checking your banking app and spot a debit card charge you do not recognize. A chat window pops up: “Hi! How can I help?” You explain. A confident reply arrives in seconds: the charge has been noted. You go to bed relieved.
Here is the problem. In the US, your bank may require you to report an error at the phone number or address it disclosed to you, so a chat message may not count (CFPB, n.d.-a). And the clock you are on is usually 60 days from your statement. It felt like talking to the bank. Legally, it may not have been.
So here is my proposition: a bank or insurer chatbot can say almost anything, but it can promise you very little, and the one thing a misplaced trust in it can cost you is time.
This is general information, not legal or financial advice. Rules differ by country and by product.
A front door, not a decision-maker
A chatbot can explain terms, find your balance and point you to forms. It usually cannot change your contract, approve a claim, or formally log a dispute the way the law requires, unless the firm has set it up to do exactly that.
Who benefits and who pays? The firm gets a front door that answers instantly at any hour, and you get speed. But if the answer is wrong, or the chat does not count as legal notice, the risk lands on you.
The evidence room: what regulators saw
In June 2023 the US Consumer Financial Protection Bureau (CFPB) published an issue spotlight called Chatbots in consumer finance. It estimated that about 37 percent of the US population, more than 98 million people, engaged with a bank’s chatbot in 2022, and that all of the ten largest commercial banks used one (CFPB, 2023). It described customers stuck in “doom loops” of unhelpful replies, inaccurate answers, and trouble reaching a person, and warned that “deficient chatbots that prevent access to live, human support can lead to law violations” (CFPB, 2023).
Read that again: a regulator saying a chatbot blocking your way to a human can be a legal problem, not just an annoyance.
A note on status: in May 2025 the CFPB withdrew 67 guidance documents (CFPB, 2025). The chatbot spotlight is not on that list and was still online in October 2026, but it was always a research report, not a rule, and the agency’s priorities have changed. Read it as a description of the problem, not current enforcement policy.
Australia’s regulator, ASIC, looked at 624 AI uses across 23 financial firms. Customer-facing generative AI was mostly kept on a short leash, such as “pre-vetted chatbot responses,” but only 10 of the 23 had policies that mentioned telling customers when AI was used, and ASIC warned the gap between AI use and governance was likely to widen (ASIC, 2024, pp. 5, 16). In the UK, the Financial Conduct Authority says it does not plan extra AI-specific rules and will rely on existing ones, including the Consumer Duty (FCA, 2026).
Why a confident answer can still be wrong
If a chatbot tells you something wrong, whether the firm is bound by it depends on your country’s law and the facts. That question is decided by the firm’s complaints process first, then by an ombudsman, regulator or court, not by the chatbot.
Part of the reason a confident answer can be wrong lies in how many of these systems are built:
Key idea: Grounding (retrieval)
Many chatbots first retrieve passages from a library of documents, such as policy wordings, fee tables and FAQs, then write a reply from them. This is called grounding, and it means the chatbot is only as good as the documents it draws on: if the library holds last year’s fee table, or misses the exclusion in your policy, the answer can be fluent, confident and wrong. For you, that is the reason to ask “which document says that?” and to check the answer against your own contract.
The best case for the chatbot
Here is the strongest argument against me. Chatbots are not a lawless zone. ASIC found that the firms it reviewed mostly kept customer-facing generative AI on a short leash, such as pre-vetted chatbot responses (ASIC, 2024). The UK regulator says the rules it already has, including the Consumer Duty, cover AI (FCA, 2026). And your existing rights did not disappear because a firm added a chat window. For a balance or a form, a chatbot may be the fastest help you get.
All true. But notice what it concedes: the protection lives in the rules and the official channels, not in the chat.
Your rights depend on where you live
United States. Your existing rights still apply. For errors on electronic transfers, such as debit card or app payments, you generally have 60 days from the statement to report them. The bank must usually investigate within 10 business days, or up to 45 days if it provisionally credits your account. But your bank may require notice at the phone number or address it disclosed, so a chat message may not count (CFPB, n.d.-a). For credit card billing errors, the notice must be written and sent to the address in your billing rights statement within 60 days; an electronic method counts only if the card issuer says it does (CFPB, n.d.-b). For insurance, your state insurance department is the regulator. As of October 2026, 27 states plus Washington, DC and Puerto Rico had adopted the insurance commissioners’ 2023 model bulletin, which reminds insurers that decisions made or supported by AI must follow all insurance laws; California, Colorado, New York and Texas have their own guidance (NAIC, n.d.; NAIC, 2026).
United Kingdom. Under the Consumer Duty, firms must make sure customers “do not face unreasonable barriers” when they need support (FCA, n.d.-a). A chatbot that will not let you reach a person is a fair thing to complain about. For most complaints the firm has up to eight weeks to give a final response; you then have six months to take it to the Financial Ombudsman Service (FCA, n.d.-b; Financial Ombudsman Service, n.d.).
European Union. Since 2 August 2026, the AI Act requires chatbots to be designed so you are told you are dealing with an AI, unless it is obvious, at the latest at your first interaction (Regulation (EU) 2024/1689, art. 50). According to law firm Goodwin, the 2026 “digital omnibus” did not delay this, though it moved stricter rules for high-risk systems, including AI for credit scoring and for life and health insurance pricing, to December 2027 (Fillon et al., 2026). Separately, data protection law gives you rights around significant decisions based solely on automated processing, including to obtain human intervention and contest the decision (GDPR, art. 22).
Canada. Complain to your bank first; it must give a detailed written response within 56 days. FCAC, the regulator, does not resolve individual complaints but can explain the process (FCAC, 2026). If you are unhappy, the Ombudsman for Banking Services and Investments (OBSI) is free and independent; you have 180 days after the final response (OBSI, n.d.).
Australia. Complain to the firm first. If that fails, an external dispute scheme hears complaints for free; for banking, credit and insurance that is the Australian Financial Complaints Authority, AFCA (Moneysmart, n.d.).
Before you close the chat window
- Save the conversation. Download the transcript or take screenshots showing the date and time.
- Ask for a person when money, a claim or a deadline is involved, and note if you could not get one.
- Get promises in writing by email, letter or secure message from the firm, not just in the chat.
- Use the official dispute route for errors and complaints, and keep the clock in mind: some deadlines are only 60 days.
Five questions to type into the chat
- Am I talking to an AI or a person, and how do I reach a person?
- Which document or policy is this answer based on?
- Does this chat count as formal notice of my dispute or complaint? If not, where do I send it?
- Can you confirm this in writing?
- If you run a firm: could you show, today, which documents your chatbot drew on for a given answer, and who approved the last change to them?
Where to complain
- US banking and credit: CFPB complaint portal. US insurance: find your state insurance department (NAIC).
- UK: Financial Ombudsman Service; FCA guide to complaining.
- Canada: FCAC; OBSI.
- Australia: AFCA; Moneysmart guide.
- EU: your firm’s complaints process first, then your national financial ombudsman or regulator.
The verdict
The evidence supports a plain conclusion. Treat a chatbot answer as information, not a promise. Anything that involves money, a claim or a deadline belongs in an official channel, confirmed in writing. The chatbot is the front desk; your rights are filed elsewhere.
What we still don’t know
We do not know how often bank or insurer chatbots give wrong answers today; neither the CFPB nor ASIC published an error rate, and ASIC’s picture dates from December 2023. We found no public data on how many disputes involve chatbot statements. And the outcome of any dispute will depend on its facts and on local law, which this guide cannot predict.
Which leaves the question worth watching: when a chatbot confidently tells a customer something wrong and the customer relies on it, who pays for the mistake? The public record does not yet say.
For organizations. If your chatbot speaks to customers, our free field manual Operational AI Readiness shows how to keep it honest. Chapter 7, Security Controls for LLM Apps and Agent Workflows, covers boundaries on what a bot can say and do; Chapter 8, Change Management for Models, Prompts, Tools, and Retrieval, covers keeping its source documents current; and Chapter 6, Runtime Visibility and the Evidence Spine, covers keeping a record of what it told whom. Read it free, or test one workflow with the free readiness assessment.
Next step. Send this to the person in your family who sorts out bank and insurance problems by app. The five questions above take a minute to type.
FairByDesign is reader-funded. If this helped, get the free Mercury Brief or see ways to support. Support never buys editorial control.
References
Australian Securities and Investments Commission. (2024). Beware the gap: Governance arrangements in the face of AI innovation (Report 798). https://download.asic.gov.au/media/mtllqjo0/rep-798-published-29-october-2024.pdf
Bakos, V. (2026). Operational AI readiness: A practitioner guide to governing AI that is already live. FairByDesign. https://fairbydesign.org/operational-ai-readiness/
Consumer Financial Protection Bureau. (2023, June 6). Chatbots in consumer finance [Issue spotlight]. https://www.consumerfinance.gov/data-research/research-reports/chatbots-in-consumer-finance/chatbots-in-consumer-finance/
Consumer Financial Protection Bureau. (2025, May 12). Interpretive rules, policy statements, and advisory opinions; withdrawal. Federal Register, 90, 20084. https://www.govinfo.gov/content/pkg/FR-2025-05-12/html/2025-08286.htm
Consumer Financial Protection Bureau. (n.d.-a). 12 CFR 1005.11: Procedures for resolving errors (Regulation E). https://www.consumerfinance.gov/rules-policy/regulations/1005/11/
Consumer Financial Protection Bureau. (n.d.-b). 12 CFR 1026.13: Billing error resolution (Regulation Z). https://www.consumerfinance.gov/rules-policy/regulations/1026/13/
Consumer Financial Protection Bureau. (n.d.-c). Submit a complaint. https://www.consumerfinance.gov/complaint/
Fillon, M., Kontosakou, A., McCluskey, C., Scott, G., Taylor, J., Ndep, J., & Thurbon, R. (2026, August 3). Not delayed, not deferred: EU AI Act transparency obligations are now in force. Goodwin. https://www.goodwinlaw.com/en/insights/publications/2026/08/alerts-technology-dpc-eu-ai-act-transparency-obligations-now-in-force
Financial Conduct Authority. (2026). Our approach to AI (updated October 2, 2026). https://www.fca.org.uk/firms/innovation/ai-approach
Financial Conduct Authority. (n.d.-a). FCA Handbook, PRIN 2A: The Consumer Duty. https://handbook.fca.org.uk/handbook/prin2a
Financial Conduct Authority. (n.d.-b). How to complain. https://www.fca.org.uk/consumers/how-complain
Financial Consumer Agency of Canada. (2026). How to file a complaint with your bank. https://www.canada.ca/en/financial-consumer-agency/services/complaints/file-complaint-financial-institution.html
Financial Ombudsman Service. (n.d.). How to complain. https://www.financial-ombudsman.org.uk/consumers/how-to-complain
Moneysmart. (n.d.). How to complain. Australian Securities and Investments Commission. https://moneysmart.gov.au/how-to-complain
National Association of Insurance Commissioners. (n.d.). Artificial intelligence [Insurance topic]. https://content.naic.org/insurance-topics/artificial-intelligence
National Association of Insurance Commissioners. (2026). Implementation of NAIC model bulletin: Use of artificial intelligence systems by insurers [Status as of October 8, 2026]. https://content.naic.org/sites/default/files/legal-adoption-map-ai-model-bulletin.pdf
Ombudsman for Banking Services and Investments. (n.d.). How the complaint process works. https://www.obsi.ca/en/for-consumers/how-the-complaint-process-works/
Regulation (EU) 2016/679 (General Data Protection Regulation), art. 22. Consolidated text via https://gdpr-info.eu/art-22-gdpr/
Regulation (EU) 2024/1689 (Artificial Intelligence Act), art. 50 and annex III. European Commission AI Act Service Desk. https://ai-act-service-desk.ec.europa.eu/en/ai-act/article-50; https://ai-act-service-desk.ec.europa.eu/en/ai-act/annex-3




Leave a Reply
You must be logged in to post a comment.