Civil Law and Autonomous Software Agent Misrepresentation Claims in Europe

1. Introduction

Autonomous software agent misrepresentation claims arise where software acts with substantial independence—such as an AI shopping agent, autonomous sales agent, financial agent, booking agent, negotiation bot, customer-service agent, or procurement system—and communicates information that is false, incomplete, misleading, or materially inaccurate, causing another person or business to enter into a transaction or suffer loss.

Examples include:

an AI sales agent falsely stating that a product has a particular specification;

an autonomous purchasing agent representing that goods are available when they are not;

an AI travel agent giving an incorrect price or availability statement;

a financial software agent making an inaccurate representation about an investment;

an autonomous chatbot falsely stating that a company provides a guarantee;

an AI negotiating agent making an unauthorized contractual representation;

an autonomous insurance agent incorrectly describing coverage;

an AI marketplace agent presenting counterfeit goods as genuine.

There is no single European civil-law doctrine specifically called “autonomous software agent misrepresentation.” The legal analysis normally combines:

contract law;

pre-contractual liability;

fraud/misrepresentation doctrines;

consumer protection;

unfair commercial practices;

electronic-commerce rules;

data-protection law where personal data is involved;

product/service liability;

agency principles under national law;

tort/delict law.

A crucial principle is:

The fact that software generated the representation does not automatically eliminate the legal responsibility of the person or company deploying or controlling the agent.

The following case law is therefore largely analogical, because European courts have not yet developed a large body of cases specifically involving fully autonomous AI agents.

2. Meaning of an Autonomous Software Agent

An autonomous software agent is software capable of:

receiving information;

analysing data;

selecting actions;

communicating with third parties;

negotiating;

purchasing or selling;

generating representations;

performing transactions with limited or no immediate human intervention.

A simple chatbot that merely displays predetermined information is different from an autonomous agent that can:

perceive → decide → communicate → transact.

The more autonomy the software possesses, the more difficult questions arise concerning attribution.

3. The Central Legal Question

Suppose:

Company A deploys an AI sales agent.

The AI tells Customer B:

“This machine has a five-year warranty.”

Customer B purchases the machine.

The actual warranty is only one year.

Who made the representation?

Possible answers include:

the AI system;

Company A;

the software developer;

the salesperson responsible for the system;

the manufacturer;

the contracting party.

European civil law generally does not treat the software itself as an independent legal person.

Therefore, the principal legal question becomes:

To which legally responsible person or entity should the agent's representation be attributed?

4. Main Types of Misrepresentation

A. False statement of fact

Example:

“This product contains 100% recycled materials.”

When it does not.

B. Misleading statement about price

Example:

“€500 total price.”

But unavoidable charges are added later.

C. Misleading omission

The AI gives technically accurate information but deliberately or systematically omits material information.

Example:

“€20 per month”

without disclosing a mandatory €200 activation fee.

D. Misleading description of characteristics

Example:

“Waterproof.”

But the product only resists minor splashes.

E. Misrepresentation of authority

An AI agent tells a customer:

“I am authorised to conclude this contract.”

But the company did not give the agent authority to make that commitment.

F. Misrepresentation caused by hallucination

The AI invents:

product specifications;

guarantees;

legal rights;

availability;

discounts;

delivery dates;

certifications.

This creates one of the most significant emerging forms of autonomous-agent liability.

5. Relevant European Legal Framework

The principal legal sources may include:

Contract law

The relevant national civil code.

Pre-contractual liability

Duties of:

good faith;

information;

disclosure;

fair dealing.

Unfair Commercial Practices Directive

Directive 2005/29/EC regulates misleading actions and omissions in consumer transactions.

Consumer Rights Directive

Information requirements can apply to distance and online contracts.

E-Commerce rules

Rules concerning online information and intermediary/service-provider responsibility may become relevant.

GDPR

Where the agent processes personal data or makes decisions based on personal data.

Digital Services Act

Relevant where the autonomous agent operates through an online platform or intermediary service.

6. Case Law 1 — Trento Sviluppo and Centrale Adriatica

CJEU, Case C-281/12, EU:C:2013:859, 19 December 2013

This is one of the most useful authorities for autonomous-agent misrepresentation.

The case concerned a promotional commercial practice and the meaning of a misleading commercial practice under Directive 2005/29/EC.

The CJEU explained that a practice is misleading where it contains false information or is otherwise likely to deceive the average consumer and is likely to cause that consumer to take a transactional decision that they would not otherwise have taken. (Infocuria)

Application to autonomous agents

Suppose an AI sales agent states:

“Only two units remain.”

If that statement is false and induces the consumer to purchase immediately, the representation may be analysed under the same misleading-practice principles.

The fact that:

“the computer generated the statement”

does not necessarily make the commercial practice legally irrelevant.

Key principle

False information + likelihood of deception + transactional effect = potentially misleading commercial practice.

7. Case Law 2 — CHS Tour Services v Team4 Travel

CJEU, Case C-435/11, EU:C:2013:574, 19 September 2013

This case concerned a travel brochure containing false information concerning hotel exclusivity.

The CJEU held that a misleading commercial practice is unfair and prohibited; it was not necessary additionally to establish that the trader had breached professional-diligence requirements. (Infocuria)

Importance

This is particularly significant for AI systems.

Suppose a company says:

“Our AI did not intentionally lie.”

That may not answer the consumer-protection question.

The relevant inquiry can instead concern whether the representation itself was misleading and capable of affecting consumer behaviour.

Autonomous-agent application

A company may argue:

“The AI generated the statement unexpectedly.”

But consumer-protection law may focus on the effect and nature of the commercial practice, rather than requiring proof that a human employee consciously intended to deceive.

Principle

Absence of human intention does not necessarily prevent a commercial representation from being legally misleading.

8. Case Law 3 — Canal Digital Danmark

CJEU, Case C-611/14, EU:C:2016:800, 26 October 2016

The case concerned advertising for satellite television.

The advertised monthly price did not adequately disclose another mandatory six-monthly charge, creating an issue under the rules concerning misleading actions and omissions. (Infocuria)

Principle

Information may be misleading not only because something false is said, but also because material information is omitted or presented insufficiently clearly.

Autonomous-agent application

Consider an AI travel agent:

“Flight €100.”

But the customer must pay:

€40 booking fee;

€25 mandatory tax;

€15 compulsory service charge.

The AI has not necessarily made a literally false statement, but the overall communication may still be misleading.

Key lesson

AI misrepresentation can occur through omission as well as through hallucinated statements.

9. Case Law 4 — Ving Sverige

CJEU, Case C-122/10, EU:C:2011:299, 12 May 2011

This case concerned advertising by Ving Sverige and the meaning of an “invitation to purchase” under the Unfair Commercial Practices Directive.

The Court considered what information must be provided when a trader presents a product and its price in a commercial communication. (Infocuria)

Autonomous-agent relevance

Imagine an autonomous shopping agent communicating:

“Buy this laptop for €799.”

The agent's communication can potentially constitute an invitation to purchase.

The legal analysis may then concern whether sufficient information has been provided about:

product characteristics;

price;

availability;

trader identity;

relevant limitations.

Principle

The legal significance of an automated representation depends upon what the communication actually invites the consumer to do.

10. Case Law 5 — Deroo-Blanquart v Sony Europe

CJEU, Case C-310/15, EU:C:2016:633, 7 September 2016

The case involved a computer sold with pre-installed software and considered whether the combined offer constituted a misleading commercial practice where the price of individual software components was not separately stated. (Infocuria)

Relevance to autonomous agents

This case is particularly useful where an AI agent bundles:

hardware;

software;

subscriptions;

cloud services;

technical support;

AI functionality.

Suppose an AI purchasing agent tells the customer:

“Complete package: €2,000.”

The customer later discovers that important software functionality requires a separate subscription.

The issue becomes whether the presentation gave sufficient information to allow an informed transactional decision.

Principle

The overall presentation of a combined offer matters; technically accurate individual statements do not necessarily prevent the overall commercial practice from being misleading.

11. Case Law 6 — Kásler and Káslerné Rábai v OTP Jelzálogbank

CJEU, Case C-26/13, EU:C:2014:282, 30 April 2014

This case concerned foreign-currency consumer credit and the transparency of contractual terms.

The CJEU examined whether the consumer could understand the economic consequences of the contractual exchange-rate mechanism and stressed the importance of plain and intelligible contractual terms. (Infocuria)

Autonomous-agent relevance

An AI agent may explain a complex contract to a consumer.

For example:

“Your effective annual cost will be approximately 4%.”

But the actual contractual mechanism may expose the consumer to significant additional costs.

The question is therefore not merely:

“Did the AI display words that were technically understandable?”

It may also be:

“Could the consumer understand the economic consequences of the contractual arrangement?”

Principle

Transparency is substantive, not merely linguistic.

12. Case Law 7 — L'Oréal v eBay

CJEU Grand Chamber, Case C-324/09, EU:C:2011:474, 12 July 2011

This case concerned trademark infringement and the liability of an online marketplace.

The CJEU considered the position of the marketplace operator and held that courts could impose measures designed not merely to terminate infringements but also to prevent future infringements. (Infocuria)

Autonomous-agent relevance

This becomes important where an autonomous agent:

selects products;

generates product descriptions;

recommends sellers;

promotes products;

processes consumer searches.

If the system repeatedly generates or distributes misleading or unlawful commercial information, the legal position may depend partly upon the role played by the platform or service provider.

Important distinction

A platform is not automatically liable merely because unlawful information exists somewhere on it.

The nature of the provider's activity and its knowledge/control can matter.

13. Case Law 8 — Glawischnig-Piesczek v Facebook Ireland

CJEU, Case C-18/18, EU:C:2019:821, 3 October 2019

This case concerned unlawful defamatory material posted by a user and the obligations of a hosting provider.

The CJEU held that EU law did not prevent a hosting provider from being ordered to remove identical unlawful material and, in certain circumstances, equivalent material. (Infocuria)

Autonomous-agent relevance

An autonomous software agent can generate enormous quantities of information.

If the agent repeatedly generates substantially identical unlawful representations, courts may need to distinguish between:

the original speaker;

the agent's operator;

the hosting provider;

the platform;

the underlying AI developer.

The case provides an important analogy concerning the responsibility of technologically mediated information systems.

14. Direct vs Analogical Case Law

A critical point for legal research is that the above cases do not establish that an AI agent itself is legally liable.

Rather, they establish principles concerning:

misleading representations;

omissions;

consumer reliance;

commercial practices;

transparency;

platform responsibility;

online information systems.

Direct authority

There is still relatively limited European case law where the defendant's autonomous AI agent itself is the central legal issue.

Analogical authority

The existing CJEU cases provide the legal principles that can be applied to AI-agent conduct.

Therefore:

The correct legal approach is to use established rules of representation and attribution rather than inventing a separate “AI misrepresentation” tort.

15. Who Is Responsible for the Agent's Statement?

Potential defendants include:

1. Principal/company

The company deploying the agent.

2. Software developer

Where the developer's own contractual or tortious duties are engaged.

3. Platform operator

Where the platform materially participates in the transaction or information system.

4. Human employee

Where the employee configured, supervised or approved the system.

5. Seller

Where the AI acts as the seller's commercial representative.

6. Manufacturer

Where the representation concerns the manufacturer's product.

The correct defendant depends upon the legal relationship and applicable national law.

16. Agency Law

A particularly difficult issue is whether autonomous software can operate as an agent in the traditional civil-law sense.

Traditional agency usually involves:

Principal → Agent → Third Party

With autonomous software:

Principal → Software → Third Party

The software may:

negotiate;

select terms;

accept offers;

send representations;

conclude transactions.

But the software itself ordinarily does not possess independent legal personality.

Therefore, the principal's authorisation, contractual arrangements, applicable agency rules and the system's actual role become important.

17. Apparent Authority

Suppose a company's website contains an autonomous AI agent.

The agent tells the customer:

“I am authorised to give you a five-year guarantee.”

The customer reasonably relies upon that statement.

Even if the company internally limited the AI to one-year guarantees, the dispute may involve concepts analogous to:

apparent authority;

reliance;

good faith;

estoppel under systems that recognise it;

pre-contractual liability;

culpa in contrahendo.

The exact doctrine varies between European legal systems.

18. Pre-Contractual Liability

Civil-law systems commonly recognise duties arising during negotiations.

A party may incur liability where it:

provides materially false information;

conceals important information;

creates unjustified reliance;

breaks negotiations in bad faith;

induces the other party to incur expenses.

Autonomous agents make this particularly complicated.

Suppose:

AI negotiation agent repeatedly tells Company B that a transaction is almost certain.

Company B spends €100,000 preparing for the transaction.

The deal collapses because the AI never had authority to conclude it.

Potential issues include:

authority;

reliance;

good faith;

culpa in contrahendo;

unjust enrichment;

damages.

19. Hallucination as Misrepresentation

AI hallucination creates a distinctive factual pattern.

Example

Customer:

“Does this insurance policy cover flood damage?”

AI:

“Yes, all flood damage is covered.”

Actual policy:

Flood damage is excluded.

The customer purchases the policy based upon the answer.

Possible claims could concern:

misleading commercial practice;

pre-contractual information;

contractual interpretation;

negligent misstatement under applicable national law;

consumer protection;

unfair terms;

regulatory duties.

The legal question is not necessarily:

“Did the AI hallucinate?”

Instead:

“Who legally supplied the information, what duty applied, and did the representation cause legally relevant reliance and loss?”

20. Misrepresentation and Consumer Protection

Consumer law can be particularly important because the consumer often does not know:

whether the answer was AI-generated;

what data the AI used;

whether a human reviewed it;

whether the AI has authority;

whether the statement is guaranteed to be accurate.

The Unfair Commercial Practices Directive focuses on the effect of misleading commercial practices on the consumer's transactional decision.

Trento Sviluppo is especially important here. (Infocuria)

21. Misleading Omission by Autonomous Agent

An autonomous agent can mislead without saying anything technically false.

Example

AI:

“The subscription costs €10 per month.”

It does not mention:

mandatory €100 joining fee;

automatic annual renewal;

minimum 24-month term.

The reasoning in Canal Digital Danmark is useful because it addresses material information presented inadequately or omitted from advertising. (Infocuria)

22. Dynamic Misrepresentation

An unusual problem with autonomous agents is that the representation may change dynamically.

At 10:00:

“Product available.”

At 10:05:

“Product unavailable.”

At 10:10:

“Product available.”

The AI may have received changing inventory information.

This raises questions concerning:

time of representation;

source of data;

reasonable reliance;

system latency;

foreseeability;

contractual formation.

23. Autonomous Negotiation

Imagine two AI agents negotiate:

Buyer AI: €100,000.

Seller AI: €110,000.

Buyer AI: €105,000.

Seller AI: Accepted.

Was a binding contract created?

Potential issues include:

authority;

offer;

acceptance;

intention;

electronic communications;

identity;

authentication;

mistake;

system error;

attribution.

European electronic-contract rules can facilitate recognition of electronic contracting, but the precise contractual consequences remain subject to applicable national law.

24. Mistake Caused by Software

Suppose an AI purchasing agent accidentally orders:

10,000 units instead of 100.

The software's operator argues:

“The AI made the mistake.”

The seller argues:

“Your system communicated the order.”

The legal question may involve:

mistake;

attribution;

electronic contracting;

authority;

allocation of technological risk.

The fact that the error was generated automatically does not itself determine whether the contract is valid or whether damages are owed.

25. Duty of Verification

Where a business deploys an autonomous agent, it may have duties concerning:

testing;

supervision;

monitoring;

updating;

accuracy;

safeguards.

The more important the representation, the stronger the potential argument that the business should not blindly allow the system to communicate unsupported claims.

For example:

AI-generated investment guarantee

is legally more serious than:

AI-generated restaurant recommendation.

The legal duty is therefore likely to depend upon the nature and consequences of the representation.

26. Causation

A misrepresentation claim normally requires a connection between:

Representation → Reliance → Transaction/Action → Loss

For autonomous software:

AI output → human reliance → transaction → economic loss

Example:

AI states that a machine is certified.

Buyer relies on statement.

Buyer purchases machine.

Machine lacks certification.

Buyer cannot legally operate it.

Buyer suffers €50,000 loss.

The claimant must establish the legally relevant causal connection.

27. Damage

Potential damages may include:

money paid under the transaction;

difference in value;

costs of correction;

wasted expenditure;

reliance losses;

foreseeable consequential losses;

contractual losses.

Pure economic loss is particularly important because many autonomous-agent misrepresentation disputes may not involve physical injury or property damage.

The applicable remedy depends heavily upon the national civil-law system and the particular cause of action.

28. Contractual vs Tortious Liability

29. Good Faith

Good faith is particularly important in civil-law systems.

A party deploying an autonomous agent should consider whether it is acting consistently with:

honesty;

cooperation;

reasonable reliance;

disclosure;

fair dealing.

An AI agent designed to maximise sales may systematically:

exaggerate benefits;

hide disadvantages;

create artificial urgency;

omit material costs.

Such behaviour can raise serious civil-law and consumer-law concerns even if no individual human employee deliberately lied.

30. Automated Commercial Persuasion

An autonomous agent can also engage in sophisticated personalised persuasion.

For example:

AI identifies that Customer A is highly price-sensitive.

It then tells A:

“This is your last chance.”

even though the product will remain available.

Potential issues include:

misleading commercial practice;

aggressive commercial practice;

manipulation;

transparency;

unfairness;

consumer vulnerability.

The Trento Sviluppo framework is relevant because the key question can be whether the practice is likely to cause a consumer to take a transactional decision they otherwise would not have taken. (Infocuria)

31. Platform Liability

Where an autonomous agent operates through a platform, liability becomes more complicated.

Consider:

Seller → Platform → AI agent → Consumer

The platform may argue:

“We merely host information.”

The claimant may argue:

“The platform's AI actively generated and presented the representation.”

The distinction between a passive intermediary and an active participant can become important.

L'Oréal v eBay is useful here because the CJEU examined the role and liability of an online marketplace and the possibility of injunctions against operators. (Infocuria)

32. Evidence

Autonomous-agent disputes require extensive digital evidence.

Important evidence includes:

prompts;

system instructions;

conversation logs;

AI outputs;

model version;

retrieval sources;

knowledge base;

timestamps;

API records;

transaction records;

human approvals;

system configuration;

audit logs;

correction history.

The claimant may need to demonstrate:

What the AI actually said at the relevant moment.

This makes preservation of AI interaction logs increasingly important.

33. Burden of Proof

The precise burden depends on the legal cause of action and national law.

Generally, the claimant may need to establish:

representation;

falsity or misleading character;

legal duty;

reliance or transactional effect where required;

causation;

damage.

But consumer-protection legislation may modify how certain elements are assessed.

CHS Tour Services demonstrates that the legal assessment of misleading commercial practice does not necessarily turn on proving subjective human intent. (Infocuria)

34. Autonomous Agent and Trade Secrets

The defendant may argue:

“The claimant cannot see how the AI works because the model is proprietary.”

But courts may still need to determine:

what data was used;

what instructions were supplied;

how the decision was generated;

whether the output was foreseeable;

whether safeguards existed.

The solution may involve:

expert evidence;

confidentiality orders;

restricted disclosure;

inspection by experts;

logs rather than source-code disclosure.

35. Cross-Border Claims

European autonomous agents frequently operate across borders.

Example:

French consumer → German company → Irish AI provider → U.S. cloud provider.

Potential legal questions include:

Which country's law applies?

Where did the misrepresentation occur?

Where was the consumer located?

Where was the contract concluded?

Which court has jurisdiction?

Is EU consumer law applicable?

Which entity is the contracting party?

For consumer contracts, EU conflict-of-law rules can provide special protections.

36. Practical Legal Test

A court analysing an autonomous-agent misrepresentation claim can conceptually follow these steps:

Step 1 — Identify the agent

What software generated the representation?

Step 2 — Preserve the output

What exactly did it say?

Step 3 — Identify the principal

Which person or company deployed the agent?

Step 4 — Determine authority

Was the agent authorised to make the representation?

Step 5 — Determine legal duty

Was the statement subject to:

contract law?

consumer law?

tort law?

professional regulation?

Step 6 — Assess accuracy

Was the statement:

false?

incomplete?

ambiguous?

misleading?

Step 7 — Assess materiality

Would the information affect the transaction?

Step 8 — Assess reliance

Did the claimant rely upon it?

Step 9 — Establish causation

Did the representation cause the transaction or loss?

Step 10 — Determine remedy

Possible remedies include:

rescission/avoidance;

damages;

price reduction;

replacement;

correction;

injunction;

other statutory consumer remedies.

37. Core Case-Law Table

38. Important Distinction: AI Error vs Legal Misrepresentation

Not every AI error creates civil liability.

For example:

AI says a restaurant is open until 11 PM when it actually closes at 10 PM.

That may be an error, but legal liability depends upon the circumstances.

By contrast:

AI says an investment is guaranteed, causing the consumer to invest €100,000.

This is potentially much more serious.

Therefore:

AI error ≠ automatically actionable misrepresentation.

The claimant normally needs to identify the legal duty, materiality, causation and damage.

39. Autonomous Software Agent Liability Formula

A useful formula is:

Autonomous Agent Misrepresentation Liability

AI Representation + Attribution + Falsity/Misleading Character + Materiality + Reliance/Transactional Effect + Causation + Damage = Potential Civil Liability

For consumer cases:

Misleading AI Communication + Average Consumer Deception + Transactional Effect = Potential Unfair Commercial Practice

For contractual cases:

AI Statement + Authorised/Attributable Agent + Reliance + Contractual Effect = Potential Contractual Remedy

40. Exam-Ready Conclusion

Autonomous software agents create a new factual environment for a largely established body of European civil and consumer law.

The central issue is not whether the software can legally “lie” as an independent person. The more important questions are:

Who deployed the agent?

Who authorised its communications?

What exactly did it represent?

Was the representation false or materially misleading?

Was important information omitted?

Was the communication capable of influencing a transactional decision?

Did the claimant reasonably rely upon it?

Did that reliance cause legally recoverable loss?

Was the agent acting as an intermediary, representative, seller or information provider?

What remedies are available under the applicable national and EU law?

The cases of Trento Sviluppo, CHS Tour Services, Canal Digital Danmark, Ving Sverige, Deroo-Blanquart, Kásler, L'Oréal v eBay and Glawischnig-Piesczek provide a strong European legal framework for analysing these questions. (Infocuria)

One-line principle: Autonomous software may generate the representation, but European civil liability ordinarily remains focused on the legally responsible human or corporate actor, the applicable duty of truthful and transparent dealing, the misleading character of the communication, reliance or transactional effect, causation, and resulting damage.