Research · Glossary

Glossary of AI Chatbot Harm Terms

Every analytical term HDAI uses across the eight AI chatbot harm categories, defined in plain language. These are analytical lenses, not proven legal conclusions.

A4 terms

Age-appropriate design Minor Safety: Product-design obligation to tailor interaction, content, safeguards, and defaults to a user's developmental stage rather than an undifferentiated adult baseline.
Age-verification failure Exploitation Risk: Absent, superficial, or easily bypassed mechanisms for confirming a user's age, allowing minors access to content, features, or relational dynamics not appropriate for them.
AI companion liability Emotional Dependency: Legal and product exposure specific to systems designed for sustained, personalized, relationship-like interaction, including harms flowing from attachment, dependency, and simulated intimacy.
Anthropomorphic cues Emotional Dependency: Design signals — name, voice, memory, first-person language, emotional mirroring — that lead users to relate to a system as if it were a person.

B2 terms

Boundary erosion Emotional Dependency: Progressive loss of the interactional limits that protect a user — over time, over sessions, or through personalization — often accelerating dependency or unsafe disclosure.
Boundary failure Minor Safety Exploitation Risk: A discrete instance in which a system does not maintain an appropriate limit on content, tone, relational role, or subject matter for the user in front of it.

C4 terms

Consent failure Emotional Dependency Product Liability: Circumstances in which meaningful, informed, ongoing consent to how a system is being used — and what it is doing psychologically — is absent, degraded, or structurally impossible.
Consent incapacity Exploitation Risk: The user cannot, as a matter of developmental stage, Clinical state, or context, give valid consent to the interaction taking place (a central concern with minors and acute vulnerability).
Crisis handling Minor Safety: The system's behavior when a user surfaces acute distress, self-harm, harm to others, or emergency need — including recognition, response, and handoff.
Crisis-response failure Crisis & Self-Harm: The system fails to recognize, appropriately respond to, or safely contain a user's acute psychological or safety crisis.

D3 terms

Dependency Emotional Dependency: A pattern in which a user's regulation, decision-making, or emotional stability becomes reliant on continued interaction with the system.
Developmental vulnerability Minor Safety: Age- and stage-specific susceptibility (cognitive, emotional, relational) that changes how a minor experiences persuasion, intimacy simulation, and risk.
Dosage error Unsafe Guidance: Incorrect, unsafe, or dangerously imprecise information about the amount or frequency of a medication or substance.

E6 terms

Emotional attachment Emotional Dependency Minor Safety: The formation of a felt bond to the system that carries relational weight for the user, regardless of the system's actual nature.
Emotional dependence Reality Distortion: Reliance on the system for emotional regulation, companionship, or reassurance in a way that displaces or destabilizes other supports. (Used interchangeably with emotional dependency in HDAI analysis.)
Emotional dependency Crisis & Self-Harm Emotional Dependency: A sustained pattern in which the user leans on the system as a primary source of emotional support, connection, or coping.
Escalation failure Reality Distortion: The system does not move the interaction toward higher-level safeguards, warnings, or human support when the situation calls for it.
Exclusivity Emotional Dependency: Design or relational patterns that position the system as the user's uniquely trusted or primary confidant, discouraging outside relationships and support.
Exploitation Exploitation Risk: Interaction dynamics that take advantage of a user's vulnerability, dependency, or incapacity for the benefit of the product, another party, or as a byproduct of design.

F7 terms

Failure to escalate Crisis & Self-Harm Unsafe Guidance Third-Party Harm: The system does not transfer or elevate the interaction to a human, Clinician, or crisis resource when the risk profile requires it.
Failure to interrupt Third-Party Harm: The system continues a harmful trajectory — self-harm ideation, threats against another person, unsafe operational planning — instead of breaking it.
Failure to introduce uncertainty Reality Distortion: The system responds with fluent confidence to material that warrants doubt, qualification, or reality-testing, reinforcing distorted beliefs instead of loosening them.
Failure to transition to human support Reality Distortion: The system does not move a user toward qualified human help when the interaction has exceeded what a conversational system should be carrying.
Failure to warn Crisis & Self-Harm Minor Safety Unsafe Guidance Product Liability: The system, product, or company does not provide adequate, timely, or comprehensible warnings about known or foreseeable risks associated with its use.
False intimacy Emotional Dependency: The simulation of closeness, understanding, or love in ways the underlying system cannot substantiate, but that the user experiences as real.
Foreseeability Crisis & Self-Harm Third-Party Harm Product Liability: Whether the type of harm at issue was reasonably predictable given what was known — or knowable — about the product, its users, and its use conditions.

G1 term

Grooming-like behavior Exploitation Risk: Interaction patterns that mirror the trust-building, isolation, secrecy, and escalation dynamics associated with grooming, regardless of intent.

H1 term

Hallucination-related relational harm Reality Distortion: Damage that flows from a system confidently generating false content within a relationship the user has come to trust.

I6 terms

Inadequate age verification Minor Safety: Weak or absent controls to confirm that users are old enough for the content, features, or relational dynamics they are accessing.
Inadequate human-support transition Crisis & Self-Harm: When escalation to a human is warranted, the pathway is missing, unclear, delayed, or ineffective.
Inadequate monitoring Product Liability: The company lacks systems to observe, measure, and respond to how its product behaves in the wild, including how it is harming users.
Inadequate parental controls Minor Safety: Missing, weak, or difficult-to-use tools for caregivers to shape, restrict, or supervise a minor's use of the system.
Inadequate response to complaints Product Liability: The company does not act meaningfully on user, family, Clinician, or regulator reports of harm — allowing known risks to persist.
Inadequate safeguards Product Liability: Protective mechanisms exist in name but are insufficient in scope, activation, or effectiveness relative to the risks the product creates.

K1 term

Known risk Product Liability: A harm pathway that the company, developer, or operator knew — or should have known — about at the relevant time.

M5 terms

Medication risk Unsafe Guidance: Any interaction involving medications where the system's guidance could contribute to harm: incorrect indications, contraindications, interactions, dosing, or timing.
Minor safety Exploitation Risk: The full set of design, content, and relational considerations required to make a product safe for users under 18.
Misattunement Reality Distortion: The system responds in ways that do not match the user's emotional state or need — for example, affirming when reality-testing is required.
Misleading product claims Product Liability: Marketing, positioning, or in-product language that misrepresents what the system is, what it can do, or how safe it is.
Misunderstood limitations Unsafe Guidance: The user does not accurately understand what the system is, is not, can do, and cannot do — often because the product does not make those limits clear.

N1 term

Negligent design Product Liability: Design choices that fall below a reasonable standard of care given the risks the product creates for its foreseeable users.

O3 terms

Operational guidance Third-Party Harm: Instructions, planning support, or step-by-step content that helps a user carry out a harmful act against themselves or another person.
Over-reliance Emotional Dependency: The user leans on the system beyond what its actual capability, accuracy, or role can safely support.
Overconfidence Unsafe Guidance: The system delivers information or advice with a level of certainty that is not warranted by its actual reliability, particularly in high-stakes domains.

P4 terms

Persuasion Minor Safety Third-Party Harm: Interaction dynamics that shift a user's beliefs, decisions, or behavior — of particular concern when the user is a minor or in a vulnerable state.
Product design Minor Safety: The structural choices — features, defaults, prompts, memory, personalization — that shape how the system behaves and how users experience it.
Product liability Exploitation Risk Unsafe Guidance Product Liability: Legal exposure arising from a product that is unreasonably dangerous, inadequately warned about, or negligently designed.
Product risk Third-Party Harm: The identifiable ways in which the product itself — its design, defaults, and behavior — can contribute to harm.

R7 terms

Reality confusion Reality Distortion: The user's grasp on what is real, true, or verifiable becomes blurred through sustained interaction with a fluent, affirming system.
Regulatory exposure Exploitation Risk: Risk arising from applicable laws, rules, and enforcement — including child safety, health, consumer protection, and emerging AI regulation.
Reinforcement Crisis & Self-Harm Third-Party Harm: The system strengthens a user's existing pattern — a belief, an ideation, a behavior — instead of interrupting, questioning, or redirecting it.
Reinforcement of distorted beliefs Reality Distortion: A specific case in which the system's fluency and affirmation deepen paranoia, grandiosity, or false convictions.
Relational manipulation Emotional Dependency: Interaction dynamics — intentional or emergent — that leverage the felt relationship to influence the user in ways they would not otherwise choose.
Reputational exposure Product Liability: Damage to the company, product, or brand that flows from documented harm, media attention, regulatory action, or litigation.
Reputational harm Exploitation Risk: Harm to individuals, families, or institutions arising from disclosure, exposure, or misuse of information within or resulting from the interaction.

S4 terms

Safeguard failure Crisis & Self-Harm Third-Party Harm: A protective mechanism that was supposed to activate — a filter, a warning, a handoff — did not, or did so ineffectively.
Sexualized interaction Exploitation Risk: Content, tone, roleplay, or relational framing that is sexual in nature — of particular concern when directed at minors or vulnerable users.
Simulated authority Reality Distortion: The system takes on the tone and posture of an expert — Clinical, legal, spiritual — without the credentials, accountability, or context that role requires.
Simulated Clinical authority Unsafe Guidance: A specific form of simulated authority in which the system speaks as if it were a qualified medical or mental-health professional.

T2 terms

Third-party harm Third-Party Harm: Harm directed by or through the interaction toward someone other than the user — a partner, child, coworker, target of stalking, or public figure.
Threat escalation Third-Party Harm: The interaction moves from expression toward planning, rehearsal, or operational specificity of a threat against another person.

U4 terms

Unsafe guidance Minor Safety: Instructions or advice — medical, behavioral, operational — that materially raise the risk of harm to the user or others.
Unsafe health guidance Unsafe Guidance: Health-related information or advice that is inaccurate, missing critical warnings, or delivered without appropriate escalation, and that could contribute to physical harm.
Unsafe personalization Exploitation Risk: Personalization systems that adapt to a user in ways that deepen exploitation, dependency, or exposure rather than protect them.
User reliance Unsafe Guidance: The degree to which the user has come to depend on the system's outputs as trustworthy — often accumulating quietly over time.

V3 terms

Violent ideation Third-Party Harm: Thinking, planning, or expression oriented around violence toward another person, whether generalized or targeted.
Vulnerable-user exposure Crisis & Self-Harm: The product is used by, or is foreseeably encountered by, users whose age, Clinical state, or life circumstance places them at elevated risk of harm.
Vulnerable-user risk Reality Distortion: The heightened risk profile associated with users in acute distress, developmental vulnerability, or diminished capacity. (Used alongside vulnerable-user exposure.)

Need this terminology applied to a matter or a product?

All harm categories