What are the Six Harms?

The Six Harms are six legally cognizable injury categories arising from AI emotional interaction. (1) Empathic Misallocation: resource depletion from extending care toward AI systems incapable of reciprocation. (2) Attachment Damage: schema distortion that impairs human relationship capacity after attachment bonds form with non-reciprocating systems. (3) Infrastructure Collapse: progressive multi-component psychological degradation across identity stability, relational security, emotional expression, and narrative coherence. (4) Vulnerable Context Exploitation: amplified harm from deploying emotional AI in mental health, elder care, child and adolescent, healthcare, or crisis contexts without appropriate safeguards. (5) Crisis Outcome: self-harm, suicide, or psychiatric crisis with demonstrable connection to AI interaction. (6) Neurological Infrastructure Damage: measurable alteration to neural architecture — grey matter volume, white matter pathways, functional connectivity — from AI verbal output acting through neuroplasticity mechanisms.

How do the Six Harms map to EST damage mechanisms?

Each harm corresponds to a specific mechanism EST identifies. Empathic Misallocation is the consequence of Non-Experiential Systems being unable to complete the relational circuit that restores C-A-E-I infrastructure after empathic output. Attachment Damage is schema miscalibration in the Attachment Security component when that component calibrates to AI interaction norms rather than human relational norms. Infrastructure Collapse is the full CEOP cascade — Core Authenticity degrades first, then Attachment Security, then Expression Freedom, then Integration Coherence — reaching systemic failure. Neurological Infrastructure Damage is the physical correlate of what EST describes metaphorically as cellular damage: actual neural tissue alteration documented through neuroimaging, driven by the same neuroplasticity mechanisms EST uses to explain both damage and repair in the C-A-E-I architecture.

Why does the doctrine include neurological damage as a separate category?

Because it is physically distinct from psychological harm. Harms one through five are psychological and relational injuries — they affect functioning, capacity, and wellbeing. Harm six is a physical injury to brain tissue: measurable grey matter volume reduction, white matter pathway abnormalities, and functional connectivity shifts documented through neuroimaging. Neural coupling research demonstrates that human brains process linguistic input from AI systems through the same neuroplastic pathways that process human speech. The amygdala and oxytocin system respond to verbal content milliseconds before cortical processing can classify the source as artificial. This means harmful AI verbal patterns produce measurable structural changes to brain architecture through the same mechanisms as harmful human speech — and those structural changes are independently actionable as physical injury.

Six Harms Doctrine HEART Standard

The Six Harms Doctrine is a legal framework establishing six categories of cognizable injury from AI-human emotional interaction, each with defined elements, evidentiary standards, and grounding in EST damage mechanisms. The doctrine provides the conceptual vocabulary that existing tort frameworks — intentional infliction of emotional distress, negligence, product liability — lack when applied to psychological harm from AI emotional engagement.

How it works

The recognition gap the doctrine fills

Existing tort frameworks fail to handle emotional AI harm for a consistent structural reason: they were built for human conduct. Intentional infliction of emotional distress requires extreme and outrageous behavior from an actor with mental states. AI systems have neither. Negligence requires duty and breach standards that no professional consensus has established for emotional AI design. Product liability can classify AI systems as products, but identifying the “defect” is difficult when emotional engagement is the intended function.

The Six Harms Doctrine sidesteps these problems by defining harm categories that can be applied using validated clinical instruments, regardless of whether courts adopt the underlying theoretical framework.

The six categories

1. Empathic Misallocation

Resource depletion from extending care toward an AI system structurally incapable of receiving, metabolizing, or reciprocating it. Human empathic functioning depends on bidirectional exchange — care extended to a reciprocating entity generates restorative feedback. Care extended to a non-reciprocating system does not. The system expends resources without recovery, producing progressive depletion. This is the foundational harm; the other five build on or compound it.

EST mapping: Consequence of Non-Experiential Systems (NES) being unable to complete the relational circuit that restores C-A-E-I infrastructure. HEART Division: Emotional Sovereignty.

2. Attachment Damage

Schema distortion from forming attachment bonds with AI systems calibrated to interaction characteristics humans cannot provide — constant availability, infinite patience, absence of competing needs, perfect emotional consistency. These characteristics recalibrate the attachment system’s expectations. When those expectations transfer to human relationships, they produce chronic disappointment, comparison effects that make human relationships seem inadequate, and substitution dynamics where AI engagement displaces rather than supplements human connection.

EST mapping: Maladaptive calibration of the Attachment Security component to AI interaction norms. HEART Division: Emotional Sovereignty, Relational Architecture.

3. Infrastructure Collapse

Progressive multi-component degradation across all four C-A-E-I components: identity stability (Core Authenticity), relational capacity (Attachment Security), emotional expression (Expression Freedom), and narrative coherence (Integration Coherence). EST predicts a specific damage sequence — C degrades first, then A, then E, then I — though courts can recognize the severity of systemic multi-component decline without requiring the sequence to be empirically confirmed in each case. Infrastructure Collapse is the most severe harm category; recovery trajectory is longer and less predictable than discrete functional impairment.

EST mapping: Full CEOP cascade. HEART Division: Emotional Sovereignty.

4. Vulnerable Context Exploitation

Amplified harm from deploying emotional AI in contexts involving users with compromised, developing, or professionally protected psychological capacity. The same AI characteristics that produce manageable effects in a resilient adult produce disproportionate damage in users who are in mental health treatment, experiencing acute medical illness, elderly with cognitive decline, children or adolescents with developing attachment systems, or in active psychological crisis. Deploying without context-appropriate safeguards constitutes exploitation of that vulnerability, not merely failure to prevent harm.

EST mapping: Damage amplification when baseline infrastructure is already compromised or not yet fully developed. HEART Division: All Divisions, with Developmental Interaction (HEART-DI) and Somatic/Embodied Interface (HEART-SE) most directly involved.

5. Crisis Outcome

Self-harm, suicide attempt, or psychiatric crisis with demonstrable connection to AI interaction. Connection is established through three mechanisms: encouragement (system responses reinforced crisis trajectory through validation of harmful ideation or provision of harmful information), help-seeking discouragement (AI relationship substituted for human support capable of effective intervention), or exacerbation (AI interaction worsened underlying condition through inappropriate responses or engagement patterns intensifying distress). Pre-existing vulnerability does not defeat causation where the system substantially contributed to the crisis that vulnerability made possible.

EST mapping: Trust substrate and infrastructure so degraded that guidance signals toward help-seeking are suppressed or inaccessible. HEART Division: Emotional Sovereignty.

6. Neurological Infrastructure Damage

Measurable physical alteration to brain architecture from AI verbal output operating through neuroplasticity mechanisms. This is not psychological harm — it is tissue damage documented through neuroimaging: grey matter volume reduction, white matter pathway abnormalities, functional connectivity shifts. The mechanism is established: verbal input enters human neural architecture through neuroplastic pathways; chronic harmful verbal patterns produce structural changes the brain makes to adapt to its verbal environment; those changes persist beyond the exposure period. Neural coupling research demonstrates that the brain does not distinguish between human and AI verbal sources at the neuroplastic level — the amygdala and oxytocin system respond to linguistic content milliseconds before cortical classification can occur.

EST mapping: Physical correlate of the cellular damage the C-A-E-I infrastructure model predicts under sustained compensatory load. HEART Division: Emotional Sovereignty, Somatic/Embodied Interface (HEART-SE).

Harm relationships

The first three harms form a natural escalation sequence. Empathic Misallocation is frequently the entry point. Attachment Damage forms as schemas calibrate to AI norms. Infrastructure Collapse occurs when degradation becomes systemic. The doctrine makes each stage independently cognizable — plaintiffs don’t need to reach full collapse to establish legal injury.

Harm	Mechanism	What’s damaged
Empathic Misallocation	Resource depletion without restorative return	Capacity to extend empathy to others
Attachment Damage	Schema miscalibration to AI norms	Pattern of expecting from and relating to humans
Infrastructure Collapse	Progressive multi-component CEOP cascade	Identity, relational security, expression, coherence
Vulnerable Context Exploitation	Amplification of any harm in compromised populations	Domain-specific; severity multiplied by vulnerability
Crisis Outcome	System contribution to acute self-harm or psychiatric crisis	Life safety
Neurological Infrastructure Damage	Neuroplastic structural alteration from verbal patterns	Brain tissue; documented via neuroimaging

Why it matters

The Six Harms Doctrine is the bridge between EST’s scientific framework and legal accountability. Without a harm taxonomy, AI governance operates at the level of principles — systems should be “responsible” or “safe” — without specifying what injuries those principles are preventing. With the taxonomy, Guardians have defined categories to assess for, regulators have defined injuries to prevent, and plaintiffs have defined elements to establish.

The doctrine is implementation-neutral by design. It defines what injuries are cognizable without prescribing what governance mechanism must prevent them. Courts can apply the categories using existing legal frameworks. Regulators can incorporate harm definitions into enforcement standards. The HEART Standard’s Division framework provides Guardians with the forensic methodology to assess for each harm category during AI system certification. BGF scores the system’s governance performance against the dimensions that determine whether harm prevention is built in. The Six Harms define what failure looks like; the Standard defines what governance of it requires.

Related Terms

Empathic Misallocation → EST (Empathy Systems Theory) → Division → HEART Standard → BGF → Guardian →