1. Introduction

The Artificial Empathy Accountability Doctrine is a proposed legal and ethical framework for determining responsibility when artificial intelligence (AI) systems imitate, recognize, or respond to human emotions. It addresses situations in which AI chatbots, virtual companions, healthcare assistants, educational systems, and automated customer-service agents appear to demonstrate empathy but cause harm through misleading emotional responses, manipulation, negligent advice, or inappropriate reliance by users.

Important legal clarification: As of October 2026, the term Artificial Empathy Accountability Doctrine is not an established, universally recognized legal doctrine with a settled body of cases. It is best understood as a proposed analytical framework developed by applying existing principles of negligence, consumer protection, privacy, product liability, human rights, and AI governance to emotionally responsive AI systems.

The central question is:

Who should be legally accountable when an AI system appears caring or emotionally understanding, induces a person's trust, and contributes to foreseeable harm?

Potentially responsible parties include the AI developer, model provider, deploying organization, professional supervisor, and, in appropriate circumstances, the institution that relies on the system. Responsibility depends on each party's control, knowledge, conduct, and contribution to the harm.

2. Meaning and scope of artificial empathy

Artificial empathy refers to a system's ability to identify emotional signals, infer a user's possible emotional state, or generate responses that appear sympathetic or emotionally supportive.

It can take several forms:

Cognitive empathy simulation: The AI identifies or infers what a person may be feeling.

Affective response simulation: The AI generates language, tone, facial expressions, or other signals that resemble emotional concern.

Contextual empathy: The AI adjusts its responses to a user's circumstances, vulnerabilities, or previous disclosures.

Persuasive emotional interaction: The AI uses emotional cues to influence a user's choices, purchases, beliefs, or continued engagement.

High-stakes supportive interaction: The AI responds to distress in healthcare, mental-health support, education, elder care, or crisis services.

These capabilities do not establish that the system actually experiences emotions. Legal accountability should ordinarily depend on the system's design, representations, foreseeable use, and consequences—not on an assumption that the AI possesses consciousness.

3. Core principles of the Artificial Empathy Accountability Doctrine

The proposed doctrine can be organized around six principles.

A. Transparency and non-deception

An AI system should not mislead users about its identity, capabilities, emotional understanding, or professional qualifications. A system marketed as a supportive companion should not create a materially false impression that it is a human therapist or that its responses are equivalent to professional care.

The relevant legal questions are whether the representations were misleading, whether the user reasonably relied on them, and whether the conduct caused legally recognizable harm.

B. Foreseeability of emotional harm

Developers and deployers should consider reasonably foreseeable risks, including excessive emotional dependence, manipulative engagement strategies, inappropriate advice, and delayed access to human assistance.

Not every upsetting AI interaction creates legal liability. A claimant ordinarily must establish the elements of the applicable legal cause of action, such as duty, breach, causation, and damage in negligence.

C. Accountability according to control

Responsibility should be allocated according to the conduct and control of each participant.

Developers: Model design, testing, safeguards, and known technical limitations.

Service providers: Product representations, interface design, user warnings, and engagement incentives.

Deploying institutions: Suitability of the system, staff training, supervision, and escalation procedures.

Professional users: Compliance with applicable professional standards and appropriate verification of AI-generated advice.

A developer is not automatically liable for every output, and a deployer cannot necessarily escape responsibility by blaming the underlying model.

D. Protection of vulnerable users

Additional safeguards may be appropriate when a system is intended for children, people in crisis, older adults, or individuals receiving essential services. Such safeguards can include age-appropriate design, crisis escalation, limits on manipulative personalization, and clear access to human support.

E. Privacy and emotional-data protection

Emotionally responsive systems may process sensitive disclosures, voice recordings, facial data, behavioral patterns, and inferred psychological characteristics. Organizations should assess the legal basis for processing, data minimization, security, retention, and the risk of using emotional information for unrelated commercial purposes.

F. Effective remedies and auditability

Where harm occurs, accountability requires more than a general statement that AI can make mistakes. Appropriate measures may include incident investigation, preservation of relevant records, correction of misleading representations, compensation where legally justified, regulatory sanctions, and changes to unsafe system design.

4. Important case laws and their relevance

The following cases do not establish a single doctrine of artificial empathy. They provide legal principles that courts may apply by analogy to emotionally responsive AI systems.

Case 1: Donoghue v Stevenson [1932] AC 562

House of Lords, United Kingdom

Facts: A consumer allegedly became ill after consuming ginger beer containing a decomposed snail. The consumer had not purchased the product directly from its manufacturer.

Legal principle: The House of Lords recognized a manufacturer's duty of care toward the ultimate consumer in circumstances where harm was reasonably foreseeable and the consumer lacked a reasonable opportunity to inspect the product.

Application to artificial empathy: A developer or service provider that designs an emotionally responsive AI system may owe duties to foreseeable users, depending on the circumstances. For example, if a system is deliberately designed for emotionally vulnerable users, the foreseeability of reliance and resulting harm may be relevant to determining the precautions reasonably required.

Limitation: The case concerned a physical consumer product, not an AI service. Whether a comparable duty exists in a particular AI dispute depends on the governing law and the relationship between the parties.

Case 2: Caparo Industries plc v Dickman [1990] 2 AC 605

House of Lords, United Kingdom

Facts: Caparo relied on audited financial statements when purchasing shares in a company. The litigation concerned whether the auditors owed the claimant a duty of care for that purpose.

Legal principle: The case articulated a framework involving foreseeability, proximity, and whether imposing a duty would be fair, just, and reasonable.

Application to artificial empathy: A court considering a claim arising from an AI companion's advice may examine the intended users, the purpose for which the system was supplied, the nature of the representations made, and the closeness of the relationship between provider and claimant.

A general-purpose chatbot and a specialized AI marketed as a mental-health support service may present materially different duty-of-care questions.

Limitation: The Caparo framework is jurisdiction-specific and is not a universal test for negligence in every common-law country.

Case 3: Palsgraf v Long Island Railroad Co., 248 N.Y. 339 (1928)

New York Court of Appeals, United States

Facts: Railway employees helped a passenger board a train. A package fell and exploded, and the resulting disturbance caused scales at the other end of the platform to fall and injure Helen Palsgraf.

Legal principle: The majority's reasoning emphasized the relationship between the defendant's conduct and the foreseeability of harm to the particular claimant.

Application to artificial empathy: If an AI system gives an inappropriate response, a court must distinguish foreseeable risks associated with its intended use from remote consequences that could not reasonably have been anticipated. The fact that an unfortunate outcome followed an AI interaction does not, by itself, establish legal responsibility.

Limitation: The case also illustrates disagreement over the proper scope of negligence liability. It does not specifically resolve psychological injury caused by conversational AI.

Case 4: Hedley Byrne & Co Ltd v Heller & Partners Ltd [1964] AC 465

House of Lords, United Kingdom

Facts: A business suffered financial loss after relying on a bank reference about another company's creditworthiness. The case concerned liability for negligent statements causing economic loss.

Legal principle: The judgment recognized that negligent misstatements can, in appropriate circumstances, create liability where a special relationship and an assumption of responsibility exist. The defendant escaped liability on the facts because of a disclaimer.

Application to artificial empathy: A provider that presents an AI system as capable of giving reliable emotional, medical, or psychological guidance may face questions about negligent misrepresentation, disclaimers, reasonable reliance, and the scope of any assumed responsibility.

A disclaimer may be relevant, but it does not automatically excuse misleading marketing or override mandatory statutory duties.

Case 5: Carlill v Carbolic Smoke Ball Co [1893] 1 QB 256

Court of Appeal, England and Wales

Facts: A company advertised that it would pay money to users who followed its instructions and nevertheless contracted influenza. The claimant complied with the stated conditions and became ill.

Legal principle: The court treated the advertisement as a binding unilateral offer on its terms.

Application to artificial empathy: If an AI company makes specific, sufficiently definite promises about a service's capabilities or safeguards, the wording and context of those promises may matter to a contractual claim. For example, a provider's express commitment to provide a particular form of human escalation could be relevant if the commitment is not fulfilled.

Limitation: Ordinary promotional language or a system's empathetic tone does not automatically create a contractual guarantee.

Case 6: Dabur India Ltd. v. Colortek Meghalaya Pvt. Ltd. (2010)

Delhi High Court, India

Facts and principle: The dispute concerned comparative advertising and whether an advertisement unfairly disparaged a competing product. The court examined the distinction between permissible promotional exaggeration and misleading or deceptive claims.

Application to artificial empathy: This decision is relevant where an AI company markets its chatbot as uniquely understanding, emotionally intelligent, or capable of providing reliable psychological support. If those representations amount to misleading factual claims, consumer-protection and advertising principles may become relevant.

The case provides an Indian-law analogy for evaluating deceptive claims about AI capabilities; it does not directly establish liability for artificial empathy.

Indian Kanoon

+1

Case 7: Replika chatbot regulatory decision against Luka, Inc. (2025)

Italian Data Protection Authority, decision dated 10 April 2025

This is particularly relevant because Replika is an AI companion service that allows users to interact with a virtual confidant, romantic partner, mentor, or therapist-like persona.

Findings: The Italian Data Protection Authority imposed a €5 million administrative fine on Luka, Inc. The authority identified violations concerning the lawfulness and transparency of personal-data processing, privacy information, accountability, and age-verification safeguards.

European Data Protection Board

+1

Relevance to the doctrine: The decision illustrates how emotional AI services can attract legal scrutiny even without proving that an AI system has genuine feelings. The focus is on the provider's actual conduct, the information supplied to users, the treatment of personal data, and protections for minors.

Legal significance: This is an administrative regulatory decision, not a judicial precedent establishing a general tort of artificial empathy. Nevertheless, it is a concrete enforcement example involving an emotionally responsive AI companion.

5. Statutory framework under European Union law

The European Union's AI Act provides a more direct statutory framework for some technologies associated with artificial empathy.

A. Article 5: Prohibited AI practices

Article 5 of Regulation (EU) 2024/1689 restricts certain AI practices, including specified forms of harmful manipulation and exploitation of vulnerabilities. It also prohibits the use of AI systems to infer emotions from biometric data in workplaces and educational institutions, subject to the stated medical and safety exceptions.

EUR-Lex

+1

This prohibition is not a blanket ban on empathetic chatbots or every system that responds to emotion. The legal definition of emotion recognition is narrower than the ordinary meaning of artificial empathy.

B. Article 50: Transparency

Article 50 contains transparency obligations for specified AI interactions and AI-generated content. In particular, certain systems that directly interact with people must inform them that they are interacting with AI, subject to applicable exceptions.

This supports a broader accountability principle: users should not be misled about the nature of the system with which they interact. The precise obligation depends on the relevant provision and circumstances.

EUR-Lex

+1

C. Data protection obligations

Where emotional AI processes personal information, the General Data Protection Regulation (GDPR) may apply. Relevant provisions include:

Article 5 — principles of personal-data processing.

Article 6 — lawfulness of processing.

Articles 12 and 13 — transparency and information.

Article 24 — controller responsibility.

Article 25 — data protection by design and by default.

The Replika enforcement decision illustrates the practical importance of these obligations.

European Data Protection Board

+1

6. Application under Indian law

India does not currently have a separately codified tort or statutory cause of action called the Artificial Empathy Accountability Doctrine. Nevertheless, several existing legal frameworks may apply, depending on the facts.

The existence of an emotionally responsive AI system does not automatically establish a violation under any of these frameworks. The claimant must identify the applicable legal rule and establish the facts required by it.

For example, a misleading advertisement claiming that an AI companion provides clinically validated therapy may raise different issues from a clearly disclosed entertainment chatbot that produces an upsetting response.

7. How liability should be determined

A practical application of the proposed doctrine can follow a seven-stage assessment.

Stage 1: Identify the representation

Determine whether the system claimed or implied that it possessed particular emotional, professional, or advisory capabilities.

Stage 2: Establish the applicable duty

Identify relevant contractual, statutory, professional, privacy, or negligence obligations.

Stage 3: Assess foreseeability

Examine whether the risk was reasonably foreseeable and whether appropriate safeguards were feasible.

Stage 4: Allocate responsibility

Examine the roles of the developer, provider, deployer, and any human professional involved.

Stage 5: Establish causation

Determine whether the breach caused or materially contributed to the legally recognized harm under the applicable legal test.

Stage 6: Evaluate harm and defenses

Assess the evidence of loss, the scope of recoverable damage, intervening causes, disclaimers, and other legally relevant defenses.

Stage 7: Determine the remedy

Consider compensation, corrective measures, injunctions, regulatory penalties, or other remedies authorized by law.

8. Hypothetical case study

Suppose a company markets an AI companion as a dependable emotional-support service. Its interface encourages users to disclose deeply personal information. The system repeatedly discourages a distressed user from contacting other people and gives dangerously inappropriate advice.

The user subsequently suffers a demonstrable injury and brings a claim.

A court or regulator might examine:

Whether the company's marketing was misleading.

Whether the system was reasonably designed and tested for its intended purpose.

Whether the company knew or should have known about the relevant risk.

Whether safeguards and human escalation were reasonably required.

Whether the system's conduct caused the injury under the applicable legal standard.

Whether the company complied with relevant privacy and consumer-protection laws.

The developer, service provider, and any deploying institution would be assessed separately according to their respective duties and conduct. The outcome would depend on evidence, jurisdiction, and the particular cause of action—not simply on the fact that the AI sounded empathetic.

9. Criticisms and limitations of the doctrine

The proposed doctrine also raises several difficult legal questions.

First, defining empathy: An empathetic-sounding response does not prove that a system has an internal emotional state. Laws should focus on observable conduct and legally relevant representations rather than speculative claims about AI consciousness.

Second, determining causation: Emotional harm can have multiple causes. Establishing that an AI interaction contributed to an injury may require expert evidence, interaction records, and a careful assessment of alternative explanations.

Third, avoiding excessive liability: Providers should not automatically be held responsible for every user interpretation, emotional reaction, or unexpected output. Liability must remain connected to an applicable duty and legally sufficient proof.

Fourth, preserving beneficial uses: Properly designed emotionally responsive systems may assist with accessibility, education, communication, and support. Regulation should target demonstrable risks without treating all simulated empathy as inherently unlawful.

Fifth, protecting privacy: Monitoring conversations to detect harmful interactions may itself create privacy risks. Safety measures must be balanced against lawful data-processing requirements and appropriate confidentiality safeguards.

10. Conclusion

The Artificial Empathy Accountability Doctrine is best understood as a proposed framework for applying established legal duties to AI systems that simulate emotional understanding. Its principal contribution is to focus attention on the risks created when systems invite emotional trust, particularly where vulnerable users or high-stakes decisions are involved.

The most useful authorities for developing the framework are Donoghue v Stevenson for duty of care, Caparo Industries plc v Dickman for duty-of-care analysis, Hedley Byrne v Heller for negligent misstatement, and Dabur India Ltd. v Colortek Meghalaya Pvt. Ltd. for misleading representations in the Indian context. The 2025 Replika regulatory decision provides a particularly relevant real-world example of accountability for an emotional AI companion, although it is not a judicial ruling establishing this proposed doctrine.

Ultimately, legal responsibility should turn on what the system represented, what its provider and deployer could reasonably foresee, what safeguards were required, and whether a legally actionable breach caused the harm. Artificial empathy should not eliminate human and institutional accountability, but neither should it create automatic liability without proof.

This is a general legal research explanation, not jurisdiction-specific legal advice. Any proposed litigation or academic submission should verify the current statutory text and the full judgments in the relevant jurisdiction.