Civil Law and AI Chatbot Financial Advice Misrepresentation in Europe

1. Introduction

AI chatbot financial advice misrepresentation arises when an AI system gives a consumer or investor inaccurate, incomplete, misleading, outdated, or unsuitable financial information, and the person relies on that information and suffers financial loss.

The important legal point is that the chatbot itself normally does not become the civil-law defendant. Liability will generally have to be attributed to a legal person—such as a bank, investment firm, financial adviser, fintech company, chatbot operator, deployer, or potentially another party in the AI supply chain.

European regulation is increasingly important here. ESMA has specifically stated that firms using AI in investment services remain subject to MiFID II requirements, including organisational duties, conduct-of-business rules and the obligation to act in the client's best interests. ESMA also identifies algorithmic bias, data-quality problems, opacity and over-reliance on AI as risks. (ESMA)

There is not yet a large body of reported European case law dealing specifically with an AI chatbot giving investment advice. Therefore, the most useful approach is to apply established European case law on investment advice, negligent misstatement, information duties, suitability, investor protection and contractual liability to the AI-chatbot setting.

2. Meaning of AI Chatbot Financial Advice Misrepresentation

A chatbot may produce statements such as:

“This investment is suitable for your risk profile.”

“This bond is guaranteed.”

“The investment has very little downside.”

“You can expect a 10% annual return.”

“This product is regulated and protected.”

“There is no significant risk of losing your capital.”

“Based on your circumstances, you should invest 40% of your savings.”

The legal problem becomes more serious when the chatbot is presented as a financial-advice service rather than merely an information tool.

A distinction should therefore be made between:

A. General financial information

Example:

“A bond is generally a debt instrument issued by a company or government.”

This normally presents a lower liability risk.

B. Personalised financial advice

Example:

“Given your income, age and risk tolerance, you should invest €20,000 in this particular fund.”

This may trigger substantially stronger duties because the communication resembles regulated investment advice.

C. Misrepresentation

This occurs where material information is:

false;

materially incomplete;

misleading;

presented with unjustified certainty;

based on outdated data;

inconsistent with the actual product;

generated despite inadequate customer information; or

accompanied by an inaccurate representation about risk.

3. European Legal Framework

Several legal regimes can operate simultaneously.

3.1 MiFID II

Where the chatbot is used by an investment firm to provide investment services, MiFID II is central.

The relevant concepts include:

suitability;

appropriateness;

customer information;

organisational requirements;

conflicts of interest;

best-interest obligations;

record keeping;

risk disclosure;

product governance.

ESMA's 2024 statement specifically confirms that AI use does not remove these existing obligations. (ESMA)

Therefore:

“The AI gave the answer” is generally not a complete defence for a regulated financial institution.

3.2 AI Act

The EU AI Act, Regulation (EU) 2024/1689, provides an additional layer of AI governance.

Its importance for financial advice includes:

transparency;

risk management;

human oversight;

documentation;

accuracy;

cybersecurity;

governance of AI systems.

The precise classification depends on the AI system and its use. The AI Act therefore should not automatically be treated as creating a private damages claim for every erroneous chatbot answer.

Instead, it operates alongside existing financial-services and private-law liability rules.

4. Main Civil-Law Causes of Action

A claimant could potentially rely on several overlapping grounds.

4.1 Breach of contract

Where a bank or fintech has undertaken to provide advisory services, incorrect advice may constitute breach of the advisory agreement.

4.2 Pre-contractual liability

A misleading chatbot interaction occurring before the investment contract may potentially generate liability where national law recognises pre-contractual duties of information and good faith.

4.3 Negligent misrepresentation

The claimant may argue that:

information was supplied;

the defendant owed a duty of care;

the information was inaccurate or inadequately verified;

reliance was foreseeable;

the claimant relied on it; and

financial loss resulted.

4.4 Tort/delict

Civil-law jurisdictions may recognise liability for negligent conduct causing economic loss, although the requirements differ considerably between jurisdictions.

4.5 Regulatory breach

Violation of financial-services rules may support a civil claim depending on the national law governing private enforcement.

5. Case Law

Case 1 — Genil 48 SL and Comercial Hostelera de Grandes Vinos v Bankinter SA and Banco Bilbao Vizcaya Argentaria SA

CJEU, Case C-604/11, ECLI:EU:C:2013:344

This is one of the most relevant EU authorities.

The case concerned interest-rate swaps and the MiFID rules governing investment services. The CJEU considered the obligations concerning suitability and appropriateness when investment services are provided to clients. (Infocuria)

Principle

Investment advice is not merely a matter of giving information. Where the relevant regulatory conditions are satisfied, the provider must comply with applicable investor-protection requirements.

Application to AI chatbots

Suppose a bank deploys an AI chatbot that:

collects information about a customer's finances;

asks about risk tolerance;

recommends a particular investment; and

presents the recommendation as appropriate.

A court could examine whether the bank actually complied with its regulatory advisory obligations rather than accepting the argument that the recommendation was merely an automated computer output.

Importance

AI cannot necessarily transform regulated advice into unregulated information merely by changing the delivery mechanism.

6. Case 2 — Länsförsäkringar Sak Försäkringsaktiebolag and Others

CJEU, Case C-542/16, ECLI:EU:C:2018:369

The case concerned financial advice provided in connection with insurance mediation.

The CJEU considered the relationship between financial advice and the EU regulatory framework applicable to insurance and investment services. The case is important because it demonstrates that the legal classification of the service actually being provided matters. (Infocuria)

Principle

A service should not be classified solely according to its label.

AI application

Suppose a company describes its chatbot as:

“Educational financial information.”

But the chatbot actually:

analyses the customer's circumstances;

recommends a particular investment;

identifies a specific product;

explains why the customer should purchase it; and

facilitates the transaction.

A court or regulator may examine the substance of the activity, rather than simply accepting the company's description.

Significance

This is especially important for fintech businesses attempting to avoid financial-advice obligations by calling an AI recommendation engine a “chatbot” or “information assistant.”

7. Case 3 — Banif Plus Bank v Lantos and Lantos

CJEU, Case C-312/14, ECLI:EU:C:2015:794

This case concerned foreign-currency consumer loans and whether certain transactions constituted investment services under the MiFID framework.

The CJEU concluded that the foreign-exchange transactions forming part of the relevant foreign-currency loan did not constitute an investment service for the purposes of the directive. (Infocuria)

Principle

Not every financial transaction automatically falls within the investment-services regime.

AI significance

This provides an important limitation.

A chatbot giving financial information does not automatically become a MiFID investment adviser merely because the subject concerns money.

The court would first need to establish:

what service was provided;

what product was involved;

whether the activity falls within the applicable financial-services legislation;

whether the provider was acting professionally;

and whether the chatbot was providing regulated investment advice.

Example

If an AI chatbot merely explains how foreign-exchange rates work, that is different from recommending a particular investment product after analysing the customer's circumstances.

8. Case 4 — BGH, XI ZR 316/13

German Federal Court of Justice, 20 January 2015

This is particularly useful for the AI problem because the BGH considered the duties of a bank providing investment advice concerning a risky currency swap.

The BGH held that an advisory bank has duties to provide investor-appropriate and product-appropriate advice. It emphasised factors such as:

the customer's knowledge;

experience;

investment objective;

willingness to take risk;

characteristics of the product;

and relevant risks.

The bank must communicate material risks in a comprehensible and non-minimising manner. (Bundesgerichtshof)

AI application

Imagine a chatbot tells a novice investor:

“This leveraged product is suitable for you and the risk is manageable.”

But the chatbot has failed to determine:

the investor's experience;

financial circumstances;

risk tolerance;

investment objectives; or

understanding of leverage.

The BGH's reasoning provides a strong analogue for arguing that the quality of the advisory process matters, not merely the existence of a disclaimer.

Important principle

The investor bears the risk that an appropriately explained investment subsequently performs badly.

But that is different from:

the investor bearing the risk of an improperly conducted advisory process.

9. Case 5 — BGH, XI ZR 247/12

German Federal Court of Justice, 2014

The BGH distinguished investment advice from other forms of financial advice and explained when an investment-advisory contract may arise.

The court emphasised that an investment-advisory relationship can arise implicitly where a customer approaches a financial institution for advice concerning the investment of money. (Bundesgerichtshof)

AI significance

This is important for chatbot systems because contracts are not necessarily created only through traditional paper agreements.

Consider a bank's application containing:

“Ask our AI investment adviser for a recommendation.”

A customer interacts with the chatbot and then purchases an investment.

The legal issue may become:

Was the chatbot merely an information interface, or was it part of the bank's advisory service?

The BGH's approach supports examining the actual relationship and circumstances rather than focusing solely on formal terminology.

10. Case 6 — BGH, III ZR 170/10

German Federal Court of Justice, 3 March 2011

The case involved investment advice and information concerning remuneration and commissions.

The BGH addressed circumstances in which information concerning distribution commissions could be relevant to the investor's decision and emphasised that misleading or inaccurate statements concerning such remuneration should not be made. (Bundesgerichtshof)

AI application

An AI chatbot might say:

“This fund is recommended because it is the most suitable option for you.”

But the financial institution receives substantially higher remuneration for selling that fund.

This creates a potential conflict-of-interest and disclosure issue.

The legal question would not merely be:

“Was the chatbot's mathematical calculation correct?”

It could also be:

“Was the recommendation affected by undisclosed incentives?”

11. Case 7 — BGH, II ZR 30/09

German Federal Court of Justice, 31 May 2010

The BGH dealt with liability arising from inaccurate information in an investment prospectus.

The court stated that investors must receive an accurate, comprehensible and complete picture of the investment, including material disadvantages and risks. (Bundesgerichtshof)

AI application

The same reasoning is highly relevant where AI generates investment information.

Suppose a chatbot describes an investment as:

“Low risk with stable returns.”

But the actual product involves:

significant leverage;

capital-loss risk;

liquidity restrictions;

complex fees; or

substantial counterparty risk.

The issue is whether the information given to the investor created a materially misleading picture of the investment.

12. Case 8 — BGH, XI ZB 2/24

German Federal Court of Justice, 11 March 2025

This case concerns investor claims involving allegedly incorrect or incomplete investment-prospectus information and pre-contractual information duties.

The BGH recognised that, in appropriate circumstances, liability for incorrect prospectus information can coexist with contractual/pre-contractual duties where the responsible persons created a particular relationship of trust with investors. (Bundesgerichtshof)

AI significance

This provides an important conceptual analogy.

If a financial institution:

develops the chatbot;

controls the investment database;

markets it as an advisory service;

supervises the chatbot;

uses it to obtain customers; and

knows customers rely on its recommendations,

the institution may face arguments concerning responsibility for the information system, notwithstanding that an algorithm generated the particular sentence.

13. Case 9 — BGH, III ZR 105/05

German Federal Court of Justice, 19 January 2006

The BGH dealt with claims arising from defective investment advice and the operation of an investment-services business without the required authorisation.

The case is relevant to the proposition that financial-advice liability can intersect with regulatory requirements. (Bundesgerichtshof)

AI application

A fintech cannot necessarily avoid regulatory or civil consequences simply because the advice is generated automatically.

Questions may include:

Was the provider authorised?

Was the service properly described?

Was the customer given accurate information?

Was the chatbot acting on behalf of a regulated firm?

Who supervised it?

Were regulatory obligations outsourced to a technology supplier?

14. Case 10 — Banif Plus Bank v Csipai and Csipai

CJEU, Case C-472/11, ECLI:EU:C:2013:88

Although this case concerned unfair contractual terms rather than AI advice directly, it illustrates the broader EU principle of effective consumer protection.

The CJEU addressed the role of national courts in examining potentially unfair terms in consumer contracts. (Infocuria)

AI relevance

AI financial platforms commonly use:

standard terms;

automated disclaimers;

limitation-of-liability clauses;

arbitration provisions;

unilateral modification clauses;

automated consent mechanisms.

A provider therefore cannot assume that a broad disclaimer automatically eliminates consumer-law scrutiny.

15. Who May Be Liable?

A chatbot dispute can involve several defendants.

A. Bank or investment firm

Usually the most obvious defendant where the chatbot forms part of its regulated advisory service.

Potential grounds:

breach of advisory contract;

MiFID-related duties;

negligent advice;

inadequate supervision;

inadequate risk controls.

B. Fintech platform

A fintech may be liable where it directly provides financial recommendations.

Questions include:

Did it present itself as an adviser?

Did it receive remuneration?

Did it select products?

Did it analyse customer circumstances?

Did it facilitate transactions?

C. AI developer

Liability of the underlying developer is more complicated.

Potential issues include:

defective model;

inadequate testing;

foreseeable misuse;

inadequate documentation;

failure to provide appropriate safeguards;

contractual allocation of responsibility.

The existence of an AI error alone does not automatically establish developer liability.

D. Financial institution using a third-party AI

This is especially important.

A bank might purchase an AI system from another company.

The customer may nevertheless argue:

“My relationship was with the bank, not with your technology supplier.”

The EBA has previously identified precisely this problem in automated financial advice: customers may direct their complaint toward the entity with which they have the customer relationship, while contractual allocation of responsibility between firms may be a separate matter. (European Banking Authority)

16. Misrepresentation Categories

16.1 False factual statement

Example:

“This investment is capital guaranteed.”

If no guarantee exists, this is relatively straightforward factual misrepresentation.

16.2 Incomplete information

Example:

“The fund historically returned 8%.”

But the chatbot fails to mention:

substantial volatility;

periods of severe losses;

fees;

illiquidity.

An incomplete statement may create a misleading overall impression.

16.3 Incorrect prediction

Example:

“This stock will probably increase by 20%.”

Predictions are more complicated.

A failed prediction is not automatically a misrepresentation because investment outcomes are inherently uncertain.

The important question is whether the chatbot presented speculation as:

fact;

certainty;

reliable professional advice; or

a properly qualified forecast.

17. Hallucination by the Financial Chatbot

AI hallucination creates a distinctive legal problem.

The chatbot might invent:

a financial regulation;

an investment product;

a tax rule;

a guarantee;

historical performance;

an interest rate;

an exemption;

a regulatory approval.

For example:

“This investment is protected by the EU investor compensation scheme.”

If that statement is false, the issue is considerably more serious than an ordinary prediction error.

18. Duty to Verify AI Outputs

A financial institution deploying AI should consider:

Data accuracy

Is the chatbot using current information?

Product database

Does it contain the correct:

fees;

risk ratings;

terms;

maturity dates;

eligibility criteria?

Model testing

Has the provider tested the system for predictable financial hallucinations?

Human oversight

Can a human adviser intervene?

Escalation

Does the system recognise when it should stop giving advice?

Audit trails

Can the institution reconstruct what the chatbot told the customer?

These questions are particularly significant because ESMA has warned about AI-specific risks including data quality, algorithmic bias and excessive reliance on AI by firms and clients. (ESMA)

19. The Role of Disclaimers

A chatbot may contain:

“This information is for educational purposes only and is not financial advice.”

That disclaimer can be relevant, but it does not necessarily determine the legal classification of the service.

The court may look at the actual conduct.

For example:

Disclaimer

“For educational purposes only.”

Actual chatbot conduct

“Based on your age, income and risk tolerance, you should invest €50,000 in Fund X.”

The second statement looks considerably more like personalised advice.

Therefore, the legal analysis should consider the substance of the interaction.

20. Causation

A claimant must generally connect the chatbot's statement to the financial loss.

The claimant may need to establish:

AI statement → reliance → investment → loss

For example:

chatbot recommended Product A;

investor relied on recommendation;

investor purchased Product A;

chatbot had incorrectly described its risk;

product subsequently lost €30,000.

The defendant may argue that the loss resulted from:

market movements;

independent investment decisions;

information from other sources;

the investor's own risk-taking;

intervening events.

Causation can therefore be one of the hardest parts of the litigation.

21. Reliance

Reliance is especially interesting in AI cases.

A claimant may demonstrate reliance through:

chatbot logs;

account records;

timestamps;

subsequent purchase orders;

messages;

transaction history;

internal platform records.

This makes digital evidence extremely important.

22. AI Chat Logs as Evidence

The claimant should ideally preserve:

the complete conversation;

prompts;

chatbot responses;

warnings;

disclaimers;

product recommendations;

timestamps;

account information;

transaction records.

The defendant may need to preserve:

model version;

system prompts;

retrieval sources;

training or fine-tuning information where legally relevant;

product database;

safety filters;

risk-classification logic;

human intervention records.

A major litigation question may therefore be:

What exactly did the AI system know and why did it produce that particular recommendation?

23. Human Oversight

Human oversight is particularly important where:

the investment is complex;

the customer is vulnerable;

the transaction is high-value;

the chatbot detects uncertainty;

the product involves leverage;

the customer requests personalised advice;

the system lacks sufficient information.

A financial institution that allows the AI to operate entirely autonomously may face greater questions about its governance and supervision.

24. Developer vs Deployer Liability

This distinction is fundamental.

Developer

Creates the AI technology.

Deployer

Uses the AI system in its business.

For example:

AI developer → Bank → Customer

The customer may interact only with the bank.

Consequently, the customer may first pursue the bank, while the bank subsequently seeks contractual indemnification from the AI developer.

This is consistent with the EBA's observation that responsibility can become complicated where different entities are involved in automated financial advice. (European Banking Authority)

25. Contractual Liability

An advisory contract may contain duties such as:

suitability assessment;

accurate information;

disclosure;

risk warnings;

conflict management;

confidentiality;

record keeping.

If the chatbot forms part of the contractual service, an inaccurate recommendation could potentially constitute a contractual breach.

The exact result depends on the applicable national law.

26. Pre-Contractual Liability

Civil-law systems frequently recognise duties arising during negotiations.

An AI chatbot could create a pre-contractual issue where:

the customer is deciding whether to invest;

the chatbot provides materially incorrect information;

the provider knows the information will influence the decision;

the customer relies on it before entering the investment contract.

German investment case law is particularly useful here because the BGH has recognised circumstances in which inaccurate investment information can generate liability based on pre-contractual duties. (Bundesgerichtshof)

27. Product Suitability

A sophisticated chatbot should not simply ask:

“How much money do you want to invest?”

It may need to consider, depending on the service and applicable law:

investment objective;

experience;

knowledge;

risk tolerance;

financial capacity;

investment horizon.

The German BGH's investment-advice jurisprudence provides a strong illustration of this principle. (Bundesgerichtshof)

28. Conflicts of Interest

Suppose an AI recommends Fund A.

But:

Fund A pays the bank a high commission;

Fund B is cheaper for the customer;

the chatbot has been configured to prioritise products generating greater revenue.

The problem is no longer simply an “AI mistake.”

It may become a question of:

conflict of interest;

misleading advice;

inadequate disclosure;

breach of best-interest obligations.

29. Damages

Depending on the applicable European national law, potential remedies may include:

A. Expectation or reliance damages

Compensation for loss caused by reliance on the incorrect advice.

B. Transaction reversal

In some legal circumstances, the claimant may seek unwinding or restitution.

C. Interest

Compensation may include statutory or contractual interest.

D. Consequential loss

Potentially recoverable where legally recognised and sufficiently connected to the breach.

E. Regulatory remedies

Separate administrative penalties may exist even where the private-law claim fails.

30. Defences Available to the Provider

A provider may argue:

1. No advisory relationship

The chatbot only provided general educational information.

2. Adequate disclaimer

The user was clearly warned that the information was not financial advice.

3. No reasonable reliance

The customer independently decided to invest.

4. No causation

The loss resulted from market conditions rather than the chatbot's statement.

5. Correct information

The chatbot's response was accurate at the relevant time.

6. Intervening misconduct

The customer ignored risk warnings or supplied false information.

7. Limitation of liability

A contractual clause may be invoked, subject to applicable mandatory consumer and financial-services law.

31. Why AI Creates a New Evidentiary Problem

Traditional advice normally involves:

human adviser → customer

AI advice may involve:

developer → model → data → retrieval system → financial institution → chatbot interface → customer

Consequently, responsibility can become fragmented.

A court may have to determine:

Who created the statement?

Who controlled the system?

Who supplied the financial data?

Who approved the product?

Who marketed the chatbot?

Who received the customer's money?

Who had the duty to supervise the system?

Who knew—or should have known—of the error?

32. Comparative Case-Law Table

33. Practical Example

Assume a European bank operates an AI chatbot called “SmartInvest.”

A consumer provides:

age: 32;

income: €40,000;

savings: €30,000;

investment experience: low.

The chatbot recommends a leveraged derivative and says:

“This product is appropriate for you and has limited downside.”

The customer invests €20,000 and loses €15,000.

The litigation could involve:

Issue 1 — Was this investment advice?

If yes, MiFID obligations may become relevant.

Issue 2 — Was the recommendation suitable?

The investor's knowledge, experience and risk profile would matter.

Issue 3 — Was the risk accurately described?

The statement “limited downside” would need to be assessed against the actual product.

Issue 4 — Was the AI properly supervised?

The bank may need to demonstrate appropriate governance.

Issue 5 — Was the investor's reliance foreseeable?

The chatbot's presentation and the bank's marketing could be relevant.

Issue 6 — Was there causation?

The investor must connect the misleading recommendation to the investment loss.

Issue 7 — Who is responsible?

Potential parties could include the bank, fintech provider and AI developer, depending on their respective legal relationships.

34. Special Problem: “AI Is Not a Financial Adviser”

A provider cannot necessarily resolve the issue simply by saying:

“The AI is not a financial adviser.”

The court may examine what the system actually does.

If it merely explains financial concepts, the statement has considerable significance.

If it:

profiles the customer;

analyses finances;

recommends securities;

identifies a particular investment;

explains why the product is appropriate;

and facilitates purchase,

the actual functionality may be much more important than the label.

This substance-over-form analysis is consistent with the approach seen in European financial-services jurisprudence. (Infocuria)

35. Key Legal Principles

The following principles are particularly important:

AI does not automatically become the legal person responsible for financial advice.

The responsible human or corporate actors remain central to civil liability.

Existing investment-advice duties can apply when AI is used as the delivery mechanism.

The substance of the service matters more than the label “chatbot.”

Personalised recommendations create greater legal risk than general information.

Suitability and appropriateness are central concepts under EU investment-services law.

Material risks must not be hidden by an AI-generated confident presentation.

False or incomplete information can create contractual or pre-contractual liability under applicable national law.

Conflicts of interest remain relevant even when recommendations are algorithmically generated.

Disclaimers do not automatically eliminate mandatory consumer or financial-services obligations.

Causation and reasonable reliance remain essential to damages claims.

AI logs and system records may become critical evidence.

Outsourcing AI technology does not necessarily transfer all customer-facing responsibility away from the financial institution.

Developer and deployer liability must be analysed separately.

National civil-law rules determine many questions concerning damages, causation, contractual liability and tort/delict.

36. Exam-Style Conclusion

AI chatbot financial advice misrepresentation in Europe represents an intersection of civil liability, financial-services regulation, consumer protection and emerging AI governance. The central legal question is not simply whether an algorithm made a mistake, but which legal person was responsible for providing the service, what duty was owed to the customer, whether the information was accurate and suitable, whether reliance was foreseeable, and whether the misinformation caused compensable loss.

The existing European jurisprudence on investment advice provides the foundation for analysing these disputes. Genil 48 demonstrates the importance of suitability and appropriateness; Länsförsäkringar illustrates the importance of correctly classifying financial services; and the German BGH investment-advice cases establish detailed principles concerning investor-specific advice, risk disclosure, conflicts and inaccurate investment information. (Infocuria)

Accordingly, the use of AI does not create a legal vacuum. In many cases, established principles of contract, pre-contractual liability, tort/delict, consumer protection and financial-services law can be applied to determine responsibility for misleading AI-generated financial advice.

Ultra-short revision formula

AI chatbot + financial recommendation + inaccurate/misleading information + reasonable reliance + financial loss = possible civil liability, subject to the applicable national law, regulatory framework, causation and proof.