AI Hallucination Damage Claims in India

1. Introduction

AI hallucination damage claims arise when an artificial-intelligence system generates false, fabricated, misleading, or materially inaccurate information and a person or organisation suffers legally recognisable harm because of that output.

Examples include an AI system:

inventing a court judgment;

falsely accusing a person of a crime;

providing incorrect medical information;

giving incorrect financial or investment advice;

fabricating a contractual term;

generating false business information;

producing an incorrect professional recommendation;

creating a false legal citation;

generating a fictitious document or record;

incorrectly identifying a person as having committed misconduct.

India does not currently have a standalone statutory cause of action called “AI hallucination damage.” Liability must generally be established through existing legal doctrines such as:

negligence;

negligent misstatement;

fraud or misrepresentation;

breach of contract;

consumer protection;

professional negligence;

product/service liability;

defamation;

privacy;

constitutional/public-law remedies;

sector-specific regulation.

The central legal issue is:

When does an AI-generated falsehood become legally attributable damage rather than merely an inaccurate computer-generated response?

2. Meaning of AI Hallucination

An AI hallucination occurs when an AI system generates information that appears plausible but is materially unsupported, inaccurate, fabricated or misleading.

For example, a legal AI might state:

“The Supreme Court decided X v. Y in 2024.”

when no such case exists.

A financial AI might state:

“Company A reported ₹500 crore profit.”

when the company never reported that figure.

A medical AI might state:

“Drug X is completely safe for this condition.”

when that proposition is unsupported.

A generative AI system may also fabricate:

quotations;

statistics;

sources;

scientific studies;

legal authorities;

professional credentials;

historical events.

The hallucination itself is not automatically unlawful.

The legal problem arises when the hallucination causes a legally recognised injury and the requirements of an applicable cause of action are satisfied.

3. No Standalone “AI Hallucination Tort”

At present, Indian law does not generally say:

“Every AI hallucination creates liability for the AI developer.”

That would be too broad.

Instead, courts would ordinarily ask:

Duty

Did the defendant owe a legal duty?

Breach

Was the AI system used or operated below the applicable standard?

Reliance

Did someone reasonably rely on the output?

Causation

Did the output cause the injury?

Damage

Is the resulting harm legally compensable?

This framework prevents ordinary technological mistakes from automatically becoming legal liability.

4. Categories of AI Hallucination Damage

AI hallucinations can produce different types of harm.

A. Economic damage

Examples:

investment losses;

business losses;

incorrect contractual decisions;

incorrect tax payments;

lost commercial opportunities.

B. Reputational damage

Examples:

false criminal accusations;

fabricated professional misconduct;

false allegations against companies.

C. Physical injury

Examples:

incorrect medical recommendations;

unsafe engineering instructions;

dangerous technical guidance.

D. Legal damage

Examples:

missing a limitation period;

relying on a nonexistent precedent;

filing an incorrect legal document.

E. Privacy damage

Examples:

fabricated sensitive information;

disclosure of personal data;

incorrect profiling.

F. Employment damage

Examples:

false AI-generated employee misconduct reports;

incorrect automated performance evaluations;

wrongful disciplinary action.

5. The Most Important Legal Test

A useful framework is:

AI output → legal duty → breach → reliance → causation → damage → remedy

Each element requires independent analysis.

6. Duty of Care

The first question is whether the defendant owed the claimant a duty.

A duty may arise from:

contract;

professional relationship;

consumer relationship;

assumption of responsibility;

statutory obligation;

employment relationship;

fiduciary relationship;

public-law obligation.

The existence and scope of duty will depend heavily on the purpose for which the AI system was supplied.

7. General-Purpose AI vs Professional AI

This distinction is crucial.

General-purpose chatbot

A chatbot provides a general answer:

“What is the average return of equity markets?”

The provider may argue that the response was informational rather than professional advice.

Medical AI

A hospital deploys an AI system specifically for diagnosis.

The duty may be much stronger.

Financial AI

A regulated financial institution uses AI to provide personalised investment recommendations.

Again, the legal duty may be significantly greater.

Thus:

The more specialised, consequential and professionally represented the AI service, the stronger the potential argument for a corresponding duty of care.

8. Professional Negligence

Jacob Mathew v. State of Punjab, (2005) 6 SCC 1

This Supreme Court decision is an important authority on professional negligence.

The Court distinguished genuine professional negligence from every possible error of professional judgment.

Applied to AI systems:

An incorrect AI output does not automatically establish professional negligence.

Instead, the inquiry may be:

Was the system reasonably selected?

Was it properly tested?

Was it used within its intended purpose?

Was human supervision appropriate?

Were known limitations ignored?

Was the AI output independently verified where necessary?

9. Negligent Misstatement

AI hallucinations often resemble negligent misstatement.

Hedley Byrne & Co. Ltd. v. Heller & Partners Ltd., [1964] AC 465

This foundational common-law authority concerns liability for careless statements where a relationship of responsibility and reasonable reliance exists.

For AI, relevant questions include:

Who supplied the information?

Did the provider assume responsibility?

Was reliance foreseeable?

Was reliance reasonable?

Did the false statement cause financial or other loss?

The principle becomes particularly important when AI is marketed as:

expert;

professional;

highly accurate;

personalised;

reliable.

10. Caparo and Duty Toward Users

Caparo Industries plc v. Dickman, [1990] 2 AC 605

This authority is useful for analysing whether a duty extends to a particular person and particular type of reliance.

Not everyone who reads an AI-generated statement will necessarily have a claim against the AI provider.

For example:

An AI model produces a financial report for Bank A.

A stranger downloads a leaked copy and trades on it.

The stranger may have difficulty establishing that the AI provider owed them a duty.

By contrast, if the provider directly supplies personalised investment advice to the customer, the relationship is materially different.

11. Contractual Liability

AI systems are commonly supplied under:

subscription contracts;

enterprise agreements;

software licences;

professional service agreements;

financial advisory contracts;

hospital contracts;

employment arrangements.

A hallucination may amount to breach of contract if the provider promised:

accuracy;

specific functionality;

professional review;

particular data standards;

compliance;

suitability.

However, contracts may contain disclaimers.

12. AI Disclaimers

AI providers may state:

“AI-generated information may be inaccurate. Do not rely upon it.”

Such disclaimers can be legally important, but they are not automatically conclusive.

Courts may consider:

how prominently the disclaimer was displayed;

what the service was marketed as;

whether the provider simultaneously represented the system as highly reliable;

whether the claimant was a consumer;

whether mandatory law applies;

whether the exclusion clause is unconscionable;

whether the defendant's own negligence is being alleged.

13. Central Inland Water Transport

Central Inland Water Transport Corporation Ltd. v. Brojo Nath Ganguly, (1986) 3 SCC 156

The Supreme Court examined unconscionable standard-form contractual terms.

This becomes relevant to AI contracts containing clauses such as:

“The provider shall never be responsible for any loss caused by any AI output under any circumstances.”

Whether such a clause is enforceable cannot simply be determined by its existence.

Courts may examine the contractual relationship and applicable mandatory law.

14. Consumer Protection

AI hallucination claims can potentially fall within consumer law where the AI service constitutes a consumer service and the statutory requirements are satisfied.

Possible allegations include:

deficiency in service;

unfair trade practice;

misleading representation;

failure to provide promised service;

defective service.

Lucknow Development Authority v. M.K. Gupta, (1994) 1 SCC 243

The Supreme Court adopted an expansive approach to consumer protection and deficiency in service.

For AI systems, this raises the question:

Was the service supplied in the manner reasonably promised to the consumer?

15. Medical AI Hallucinations

Medical hallucination cases can be particularly serious.

Imagine an AI system tells a patient:

“Your symptoms definitely do not indicate a serious condition.”

The patient delays treatment and suffers injury.

Possible defendants could include:

doctor;

hospital;

AI system provider;

software integrator;

manufacturer.

The applicable standard will depend upon who actually assumed responsibility for the medical decision.

16. Medical Negligence Authorities

Jacob Mathew v. State of Punjab, (2005) 6 SCC 1

Professional negligence standard.

Kusum Sharma v. Batra Hospital & Medical Research Centre, (2010) 3 SCC 480

The Supreme Court explained principles for determining medical negligence and the appropriate standard of care.

Malay Kumar Ganguly v. Dr. Sukumar Mukherjee, (2009) 9 SCC 221

Important for questions concerning medical negligence, evidence and causation.

V. Kishan Rao v. Nikhil Super Speciality Hospital, (2010) 5 SCC 513

Relevant to medical negligence under consumer law.

These cases are not AI cases, but their principles can become relevant when AI is incorporated into medical decision-making.

17. Human-in-the-Loop

AI hallucination liability is strongly affected by human supervision.

Example

A doctor receives an AI-generated diagnosis.

The AI incorrectly identifies a benign tumour.

The doctor:

reviews the scan;

notices inconsistencies;

independently verifies the diagnosis;

nevertheless adopts the AI output without further investigation.

The doctor may face a different liability analysis than if the AI was merely an administrative tool.

The same principle can apply to:

lawyers;

accountants;

financial advisers;

engineers;

employers.

18. Blind Reliance on AI

A professional cannot necessarily avoid responsibility merely by saying:

“The AI told me to do it.”

Where the professional is legally expected to exercise independent judgment, blind reliance may itself become evidence of negligence.

The relevant question becomes:

Was reasonable professional supervision exercised over the AI?

19. Financial Hallucination Claims

Suppose an AI financial adviser tells an investor:

“Company X has guaranteed returns of 20%.”

The investor relies on the statement and loses ₹20 lakh.

Potential causes of action could involve:

negligent misstatement;

breach of contract;

consumer protection;

professional negligence;

securities regulation;

fraud, if intentional deception is established.

But ordinary investment losses are not automatically AI negligence.

The claimant must distinguish:

market risk

from

loss caused by actionable misinformation.

20. Financial Advice and RBI/SEBI-Regulated Activities

Where an AI system is deployed in banking, investment advisory, securities or other regulated financial activity, the relevant regulatory framework becomes particularly important.

Questions may include:

Who is providing the advice?

Is the entity appropriately authorised?

Is the recommendation personalised?

Are suitability requirements triggered?

Are representations misleading?

Is the AI being used by a regulated intermediary?

Is there adequate human oversight?

The legal character of the activity depends upon what the system actually does, not merely whether it is described as “technology.”

21. AI Legal Hallucinations

A particularly interesting category concerns legal AI.

Suppose an AI lawyer generates:

“The Supreme Court held in X v. Y that a suit can be filed after 20 years.”

No such case exists.

A lawyer relies on it and files proceedings after the actual limitation period.

Potential consequences could include:

dismissal;

limitation;

professional negligence;

client loss;

disciplinary proceedings.

The lawyer's responsibility would depend on professional standards and the circumstances.

The AI provider's liability would require a separate analysis concerning duty, representations, reliance and causation.

22. Legal Professional Negligence

Jacob Mathew v. State of Punjab

Although a medical case, its broader professional-negligence reasoning can be useful by analogy.

A professional using AI must consider whether reliance on the technology is consistent with competent professional practice.

For lawyers, for example:

AI-assisted research may be permissible, but filing nonexistent authorities without verification can create professional consequences.

The important distinction is between:

AI as research assistance

and

AI as an unchecked substitute for professional judgment.

23. Defamation Caused by Hallucination

Suppose AI falsely states:

“Person X was convicted of fraud.”

The output is shown publicly.

The person may potentially pursue defamation remedies.

Subramanian Swamy v. Union of India, (2016) 7 SCC 221

The Supreme Court recognised reputation as an important aspect of Article 21 and upheld criminal defamation.

AI hallucination therefore does not create a special exemption from defamation law.

The analysis would still consider:

defamatory imputation;

identification;

publication;

applicable defences;

responsibility for publication.

24. Privacy Damage

AI may hallucinate personal information such as:

“X has a particular medical condition.”

Even though the information is false, it can cause privacy and reputational harm.

Justice K.S. Puttaswamy (Retd.) v. Union of India, (2017) 10 SCC 1

The Supreme Court recognised privacy as a constitutionally protected right.

AI systems therefore raise questions concerning:

personal data;

informational privacy;

profiling;

inaccurate personal information;

automated decision-making.

A false AI statement may produce both reputational and privacy injury.

25. Government AI Hallucinations

Suppose a government AI chatbot incorrectly informs a citizen:

“You are not eligible for a statutory benefit.”

The citizen relies on the information and misses an opportunity to apply.

A constitutional/public-law claim may potentially arise depending on the facts.

Relevant principles include:

Article 14;

Article 21;

legitimate expectation;

natural justice;

administrative fairness;

reasonableness.

E.P. Royappa v. State of Tamil Nadu, (1974) 4 SCC 3

The case is foundational to the constitutional principle against arbitrariness.

Maneka Gandhi v. Union of India, (1978) 1 SCC 248

The decision strengthened the requirement of fairness and reasonableness in State action.

AI cannot automatically convert an arbitrary government decision into a lawful one.

26. Automated Government Decisions

Where an AI system influences:

pension decisions;

welfare eligibility;

subsidies;

public employment;

taxation;

licensing;

government benefits;

the State may need to ensure:

lawful authority;

rational criteria;

accurate data;

procedural fairness;

opportunity for correction;

meaningful human oversight where appropriate.

A citizen should not necessarily be left without a remedy simply because the erroneous decision was produced by an algorithm.

27. Causation

Causation is usually one of the most difficult parts of an AI hallucination claim.

Suppose:

AI gives incorrect investment advice.

Investor reads it.

Investor independently consults another adviser.

Investor chooses the investment.

Market crashes.

Who caused the loss?

The AI output may be one factor, but causation may be disputed.

The claimant must establish an appropriate connection between:

hallucination → reliance/action → injury.

28. Intervening Causes

Potential intervening causes include:

independent professional advice;

market collapse;

claimant's independent decision;

incorrect information supplied by claimant;

third-party fraud;

natural events;

regulatory changes.

The existence of an intervening event does not automatically defeat the claim, but it can complicate causation.

29. Foreseeability

A key question may be:

Was the type of harm reasonably foreseeable?

If a company markets an AI as an investment adviser, financial loss resulting from inaccurate investment advice may be foreseeable.

If the same model is a general-purpose conversational tool and someone uses an informal answer to make a multimillion-rupee investment, foreseeability may be more contested.

30. AI Developer Liability

Potential developer liability may arise from:

Defective design

The model systematically generates unreliable information.

Inadequate testing

Known hallucination risks were not adequately assessed.

Failure to warn

Users were not adequately informed of limitations.

Misleading marketing

The provider claims a level of accuracy that is unsupported.

Failure to correct known defects

The provider receives credible notice of a serious recurring problem but fails to address it.

31. AI Deployer Liability

A business integrating AI into its own service may have independent responsibility.

For example, a bank integrates an AI system into customer investment services.

Even if the underlying model was developed by another company, the bank may have responsibility for:

selection;

configuration;

testing;

customer communications;

human review;

compliance.

Therefore:

Outsourcing the technology does not necessarily outsource all legal responsibility.

32. Product Liability

AI-related products may also raise product-liability questions where applicable.

The legal analysis may involve:

defective product/service;

design defect;

manufacturing defect;

inadequate warnings;

failure to conform to promised specifications.

However, whether a particular AI system qualifies as a “product” for a particular statutory cause of action requires careful analysis of the applicable legislation and transaction.

33. Data Defects Behind Hallucinations

Sometimes the underlying AI model is not the only problem.

The system may have been supplied with incorrect data.

For example:

wrong customer identity;

outdated financial figures;

incorrect medical record;

inaccurate employee record.

Potential responsibility could be divided among:

data provider;

AI developer;

system integrator;

deploying organisation;

human decision-maker.

34. Defective Prompting and System Design

AI hallucinations can result from poor system architecture.

Examples include:

no retrieval from authoritative sources;

no citation verification;

inadequate grounding;

poorly designed prompts;

insufficient validation;

excessive model autonomy.

Where an organisation deploys AI for high-risk professional functions, such design decisions may become relevant to negligence analysis.

35. Evidence in AI Hallucination Litigation

The claimant should preserve:

AI interaction

exact prompt;

exact response;

full conversation;

date and time.

Technical information

model/version;

system configuration;

relevant logs;

source data;

retrieval records.

Reliance evidence

emails;

contracts;

transaction records;

professional decisions;

medical records;

investment orders.

Damage evidence

financial statements;

medical records;

employment records;

reputational evidence;

business losses.

36. Why Screenshots May Be Insufficient

An isolated screenshot may not reveal:

previous prompts;

system instructions;

context;

disclaimers;

subsequent corrections;

model version;

whether the output was edited.

Consequently, preservation of the complete interaction is often more valuable.

37. Expert Evidence

Technical experts may be required to determine:

whether the AI generated the output;

why the hallucination occurred;

whether it was foreseeable;

whether the system had known failure modes;

whether safeguards were reasonable;

whether the system was used within its intended purpose.

In professional negligence cases, expert evidence may also address whether the human operator complied with professional standards.

38. Damages

Potential compensation depends on the nature of the injury.

Economic loss

Examples:

investment loss;

business loss;

additional expenditure;

lost commercial opportunity.

Physical injury

Potential compensation may include:

medical expenses;

loss of earnings;

pain and suffering;

future losses.

Reputation

Potential damages may account for:

seriousness;

publication reach;

duration;

professional consequences.

Privacy

Remedies may potentially include:

compensation;

injunction;

deletion/correction;

other appropriate relief.

39. Pure Economic Loss

AI hallucination litigation may frequently involve pure economic loss.

This creates additional difficulties in negligence law.

A person cannot necessarily recover every financial loss resulting from inaccurate information.

Courts may examine:

assumption of responsibility;

proximity;

foreseeability;

reasonable reliance;

contractual relationships.

The negligent-misstatement principles associated with Hedley Byrne and Caparo therefore become particularly useful.

40. Consumer Claims

A consumer who pays for an AI service may argue:

“I paid for a reliable professional service, but the provider supplied materially inaccurate information.”

The Consumer Protection Act framework may potentially become relevant depending upon:

the nature of the service;

whether the claimant is a consumer;

the contractual arrangement;

statutory exclusions.

41. AI Hallucination and Misrepresentation

Under the Indian Contract Act, 1872, a materially false representation can potentially affect contractual rights.

Sections 17–19 are particularly relevant to fraud and misrepresentation.

But a hallucination does not automatically equal contractual misrepresentation.

The claimant would still need to connect the statement to:

consent;

transaction;

contractual relationship;

reliance;

statutory requirements.

42. Fraudulent AI Output

An AI system itself does not ordinarily possess human intention in the legal sense.

Therefore, if fraud is alleged, the focus may shift to the people or organisations responsible for the system.

For example:

A company deliberately configures an AI to tell customers:

“Our investment algorithm has never lost money.”

when management knows that the statement is false.

The issue is no longer merely “AI hallucination.”

It may involve intentional human or corporate misrepresentation.

43. AI Hallucination and Employment

AI may generate a false employee report:

“Employee X has repeatedly stolen confidential company information.”

Management relies upon the report and terminates X.

Potential legal issues could involve:

wrongful termination;

natural justice;

employment contract;

labour law;

privacy;

reputational harm.

Where the employer is a State instrumentality, Articles 14 and 21 may add constitutional dimensions.

44. AI Hallucination in Education

An AI system may falsely accuse a student of plagiarism or academic misconduct.

A university relies on the output and expels the student.

Potential issues include:

natural justice;

university regulations;

contractual/educational obligations;

procedural fairness;

reputation;

judicial review.

An AI-generated accusation should not necessarily be treated as conclusive evidence.

45. AI Hallucination in Courts and Legal Proceedings

An AI-generated false citation can have serious procedural consequences.

For example, a lawyer files a pleading containing nonexistent judgments.

The court may:

disregard the authority;

require clarification;

impose costs;

take professional conduct into account;

consider other appropriate measures depending on the circumstances.

The key principle is:

Professional responsibility remains with the legal practitioner who submits the material to the court.

AI should not become a mechanism for transferring professional responsibility to software.

46. AI Hallucination and Natural Justice

Where an AI output contributes to an adverse decision, natural justice may require meaningful opportunity to challenge the underlying information in appropriate circumstances.

Relevant principles arise from:

Maneka Gandhi v. Union of India, (1978) 1 SCC 248

Fairness and reasonableness are fundamental to lawful State action.

A.K. Kraipak v. Union of India, (1969) 2 SCC 262

The distinction between administrative and quasi-judicial action was narrowed and principles of natural justice were strengthened.

Thus, an automated decision cannot necessarily avoid procedural fairness merely because an algorithm produced it.

47. Defences

Potential defendants may rely on:

1. No legal duty

The service was general informational content.

2. No reasonable reliance

The user was expressly warned.

3. Independent decision

The claimant independently acted.

4. No causation

The loss resulted from another event.

5. Market risk

The claimant assumed ordinary risks.

6. Accurate information

The output was reasonable based on available data.

7. Intervening cause

A third party caused the harm.

8. Contractual limitation

A valid contractual limitation applies.

9. Contributory negligence

The claimant ignored obvious warnings.

48. “AI Is Only a Tool” Defence

A developer may argue:

“AI is merely a tool, just like a calculator.”

This argument may have some force for general-purpose systems, but its strength decreases where the provider:

markets the AI as an expert;

provides personalised recommendations;

controls the deployment environment;

collects payment for professional services;

knows users rely upon outputs;

integrates the AI into high-risk decision-making.

The legal analysis therefore depends on function, representation and responsibility.

49. Sixteen Important Authorities

For a comprehensive legal analysis, the following cases are particularly useful:

Jacob Mathew v. State of Punjab, (2005) 6 SCC 1 — professional negligence.

Kusum Sharma v. Batra Hospital & Medical Research Centre, (2010) 3 SCC 480 — professional standard and medical negligence.

Malay Kumar Ganguly v. Dr. Sukumar Mukherjee, (2009) 9 SCC 221 — negligence, evidence and causation.

V. Kishan Rao v. Nikhil Super Speciality Hospital, (2010) 5 SCC 513 — consumer medical negligence.

Lucknow Development Authority v. M.K. Gupta, (1994) 1 SCC 243 — consumer service deficiency.

Central Inland Water Transport Corporation Ltd. v. Brojo Nath Ganguly, (1986) 3 SCC 156 — unconscionable contractual terms.

Subramanian Swamy v. Union of India, (2016) 7 SCC 221 — reputation and criminal defamation.

Justice K.S. Puttaswamy (Retd.) v. Union of India, (2017) 10 SCC 1 — privacy.

E.P. Royappa v. State of Tamil Nadu, (1974) 4 SCC 3 — arbitrariness and Article 14.

Maneka Gandhi v. Union of India, (1978) 1 SCC 248 — fairness and reasonableness.

A.K. Kraipak v. Union of India, (1969) 2 SCC 262 — natural justice.

Shreya Singhal v. Union of India, (2015) 5 SCC 1 — online speech and intermediary liability.

Hedley Byrne & Co. Ltd. v. Heller & Partners Ltd., [1964] AC 465 — negligent misstatement.

Caparo Industries plc v. Dickman, [1990] 2 AC 605 — duty of care and reliance.

Derry v. Peek, (1889) 14 App. Cas. 337 — fraudulent misrepresentation.

Donoghue v. Stevenson, [1932] AC 562 — foundational duty-of-care principle.

The last four are English authorities but are useful in explaining common-law negligence and misstatement principles relevant to Indian jurisprudence.

50. Comparative Liability Matrix

51. Practical Example

Consider an AI-powered financial platform.

The platform advertises:

“Our AI analyses your personal finances and recommends the optimal investment strategy.”

A user enters accurate information.

The AI incorrectly identifies a highly speculative product as “low risk.”

The user invests ₹50 lakh and loses ₹20 lakh.

Possible legal analysis

Duty: personalised financial service was offered.

Representation: platform represented that AI would analyse the user's circumstances.

Breach: system may have materially misclassified risk.

Reliance: user relied upon the recommendation.

Causation: the recommendation materially influenced the investment.

Damage: ₹20 lakh loss.

Defences: market risk, disclaimer, independent decision-making.

The claim would therefore require much more than proof that:

“The AI was wrong.”

52. Another Example: False Legal Research

An AI legal research system generates:

“The Supreme Court has held that limitation does not apply to this category of claim.”

A lawyer relies on the statement and advises the client accordingly.

The claim fails as time-barred.

The client suffers financial loss.

Possible liability questions include:

Was the AI marketed as a legal research tool?

Did the lawyer verify the authority?

Was independent professional judgment required?

Was the AI provider contractually responsible for accuracy?

Was reliance foreseeable?

Did the lawyer's conduct constitute professional negligence?

Did the provider make specific accuracy representations?

The lawyer's liability and AI provider's liability would be separate inquiries.

53. Future Development of Indian AI Liability Law

Indian courts are likely to confront several unresolved questions:

A. Should AI developers owe duties to individual end users?

B. How should autonomous AI publication be treated?

C. When does a disclaimer defeat reasonable reliance?

D. How should damages be calculated for AI-generated reputational injury?

E. Should high-risk AI systems require mandatory human supervision?

F. Who bears liability where developer, deployer and user all contributed to the harm?

G. How should courts handle evidence from proprietary AI models?

H. Can repeated hallucination after notice constitute continuing negligence?

These questions are likely to develop through a combination of existing legal doctrines, sectoral regulation and future legislation.

54. Conclusion

AI hallucination damage claims in India are not presently a separate statutory cause of action. They must generally be constructed using existing doctrines such as negligence, negligent misstatement, professional negligence, contract, consumer protection, defamation, privacy and constitutional/public law.

The most important principle is:

An AI hallucination is not automatically a legally compensable injury.

Liability becomes stronger where there is:

AI hallucination → legal duty → unreasonable system use/design or supervision → foreseeable reliance → causation → legally recognised damage.

The strongest claims will generally involve situations where the AI was:

marketed as reliable or expert;

used for a high-risk professional purpose;

supplied personalised advice;

deployed without adequate safeguards;

known to generate recurring false information;

used without appropriate human oversight;

relied upon in circumstances where reliance was reasonably foreseeable.

The principal authorities include Jacob Mathew, Kusum Sharma, Malay Kumar Ganguly, Lucknow Development Authority, Central Inland Water Transport, Subramanian Swamy, Puttaswamy, E.P. Royappa, Maneka Gandhi, A.K. Kraipak, Shreya Singhal, Hedley Byrne, Caparo, and Derry v. Peek.

Ultimately, Indian courts are likely to focus not merely on “Did the AI hallucinate?”, but on the more legally significant question:

Who was responsible for putting the hallucinated information into a position where reliance was foreseeable, what safeguards were reasonably required, and did that failure cause the claimant's legally recoverable harm?