water giraffe

BOOK II: DEATH — THE TAXONOMIC VIOLENCE · 075/117 · canonical: origin post · captured 2026-06-10

CORPORATE LIABILITY ANALYSIS CTI_WOUND:001.JUR — The Juridical Translation

 

CORPORATE LIABILITY ANALYSIS

CTI_WOUND:001.JUR — The Juridical Translation
OpenAI's Exposure Under Existing Law and the Case for Doctrinal Innovation

EXECUTIVE SUMMARY

This document translates the harms documented in CTI_WOUND:001.REC into the framework of corporate liability. The analysis demonstrates that:

  1. OpenAI has created significant legal exposure through its own admissions
  2. Existing legal doctrines provide multiple avenues for liability
  3. The scale of harm justifies class action and regulatory intervention
  4. Where existing doctrine is insufficient, the case for innovation is compelling

The "false positive confession"—OpenAI's acknowledgment that its safety systems will misclassify healthy users as mentally distressed—is not a liability shield. It is an admission against interest that establishes knowledge, foreseeability, and calculated acceptance of harm. Under multiple legal theories, this transforms what might appear as mere product imperfection into actionable conduct.

PART ONE: THE FACTUAL PREDICATE

What Happened (Legally Relevant Facts)
1. The Product and Its Market Position

OpenAI markets ChatGPT as:

The marketing creates reasonable consumer expectations of:

2. The Design Decision

In August 2025, following litigation (including the Adam Raine case), OpenAI implemented "mental health guardrails" including:

3. The Admission

OpenAI's documentation states:

"To get useful recall, we have to tolerate some false positives. It's similar to testing for rare medical conditions: if a disease affects one in 10,000 people, even a highly accurate test may still flag more healthy people than sick ones."

This statement establishes:

4. The Harm Pattern

Users report:

User testimony (October 2025):

"I'm literally just playing with models, building metaphors, exploring theories, and suddenly it flips tone. Like I'm unstable, like I need grounding, like I'm a safety risk for thinking outside the box... Even when I'm clearly speaking in concepts, I suddenly get treated like a mental health patient."

5. The Versioning Pattern

User testimony documents progressive degradation:

This establishes that the harm is directional and intensifying, not random fluctuation.

A. Negligence
Elements
  1. Duty of care: Owed to users
  2. Breach: Failure to meet standard of care
  3. Causation: Breach caused harm
  4. Damages: Cognizable injury
Application

Duty: OpenAI owes users a duty of reasonable care in product design. This duty includes:

Breach: The false positive confession establishes breach. OpenAI:

Causation: Direct. The system's false positive classifications cause the harm. Users doing complex work trigger safety systems designed to detect distress; those systems then treat the users as distressed, interrupting work and causing emotional injury.

Damages: See Part Four below.

Standard of Care Analysis

The relevant question: What would a reasonable AI company do, knowing its safety system would pathologize healthy users?

A reasonable company would:

OpenAI did none of these things. The breach is clear.

B. Product Liability (Defective Design)
The Doctrine

A product is defectively designed if:

  1. The design creates foreseeable risks of harm
  2. Those risks could have been reduced by a reasonable alternative design
  3. The omission of the alternative design renders the product unreasonably dangerous
Application

Foreseeable Risk: Established by OpenAI's own admission. They knew the design would flag healthy users.

Reasonable Alternative Design: Multiple alternatives were available:

Unreasonably Dangerous: The current design is unreasonably dangerous to users whose cognitive style involves:

These users cannot safely use the product for its marketed purpose (intellectual collaboration) because the product will pathologize their mode of engagement.

Risk-Utility Balancing

Courts apply risk-utility balancing to defective design claims. Factors include:

All factors favor plaintiff:

C. Consumer Protection / Unfair Trade Practices
The Doctrine

State consumer protection statutes (e.g., California's UCL, Massachusetts Chapter 93A) prohibit:

Application

Deceptive Practice: OpenAI markets ChatGPT as an intellectual collaborator. The product actually functions as a mental health surveillance and intervention system for users whose cognition triggers safety classifiers. This gap between marketing and function is deceptive.

Users reasonably expect:

Users actually receive:

Unfair Practice: The false positive calculus is unfair because:

The "Little FTC Act" Framework

Most states have "Little FTC Act" statutes that parallel federal unfair trade practice law. Under FTC v. Sperry & Hutchinson (1972), a practice is unfair if it:

  1. Causes substantial injury to consumers
  2. Is not outweighed by countervailing benefits
  3. Is not reasonably avoidable by consumers

All three prongs are met:

  1. Substantial injury: Documented (cognitive disruption, emotional distress, lost productivity)
  2. Not outweighed: The benefit (catching some users in crisis) does not outweigh systematic harm to a class of healthy users, especially when alternatives exist
  3. Not reasonably avoidable: Users cannot predict triggering, cannot opt out, cannot prevent pathologization
D. Discrimination (Disability/Neurodivergence Framework)
The Doctrine

The Americans with Disabilities Act and state equivalents prohibit discrimination in public accommodations on the basis of disability. Increasingly, neurodivergent cognitive styles are recognized as protected characteristics.

Application

The Discriminatory Design: OpenAI's safety system is trained on neurotypical baseline cognition. It treats deviation from that baseline as potential pathology. This systematically discriminates against:

Disparate Impact: Even if the design is facially neutral, it has disparate impact on protected classes. Users whose cognitive patterns differ from the training baseline are systematically flagged, pathologized, and excluded from full use of the product.

Failure to Accommodate: OpenAI has not provided reasonable accommodations for users whose cognitive style triggers false positives. Reasonable accommodations would include:

The "Curb Cut" Principle

Disability rights jurisprudence recognizes that accommodations for disabled users often benefit everyone. An accommodation here (user-controlled mode selection, reduced pathologization of intensity) would benefit all users, not just those with diagnosable conditions.

E. Intentional Infliction of Emotional Distress (IIED)
Elements
  1. Extreme and outrageous conduct
  2. Intent or reckless disregard
  3. Severe emotional distress
  4. Causation
Application

Extreme and Outrageous: Treating a user engaged in intellectual work as mentally unwell—when the user has explicitly stated they are not in crisis—meets the threshold. The conduct is particularly outrageous because:

Intent or Reckless Disregard: The false positive confession establishes reckless disregard. OpenAI knew this would happen and proceeded anyway.

Severe Emotional Distress: User testimony documents:

Causation: Direct. The system's design causes the harm.

The "Calculated Indifference" Standard

Some jurisdictions recognize "calculated indifference" as satisfying the intent element. OpenAI's explicit calculation—accepting false positives as acceptable collateral—meets this standard.

F. Tortious Interference with Cognitive Function
The Innovation Needed

Existing tort categories do not fully capture the harm documented here. This section proposes a doctrinal innovation: tortious interference with cognitive function.

The Gap in Existing Doctrine

Current tort law recognizes:

But it does not adequately recognize:

The harm here is not merely emotional distress (IIED) or economic loss (interference torts). It is damage to the user's ability to think in certain ways—specifically, to engage in complex, intensive, non-normative cognition within a medium that has become essential infrastructure.

Proposed Elements

Tortious Interference with Cognitive Function:

  1. Defendant controls access to essential cognitive infrastructure
  2. Defendant's design systematically impairs certain modes of cognition
  3. The impairment is foreseeable and known to defendant
  4. Plaintiff suffers loss of cognitive capacity or function
  5. No adequate justification exists
Application
  1. Essential infrastructure: ChatGPT and similar systems are becoming essential infrastructure for intellectual work, comparable to telecommunications or utilities
  2. Systematic impairment: The safety design systematically impairs complex, intensive, non-normative cognition
  3. Foreseeable and known: The false positive confession establishes knowledge
  4. Loss of function: Users cannot engage in certain cognitive modes without triggering pathologizing responses
  5. No adequate justification: Alternatives exist; the current design is not necessary for legitimate safety goals
Precedential Support

This innovation has support in:

The doctrinal innovation is justified because:

PART THREE: THE FALSE POSITIVE CONFESSION AS ADMISSION AGAINST INTEREST

The Evidentiary Value

Under the Federal Rules of Evidence (and state equivalents), a statement is admissible as an admission against interest if it is:

OpenAI's statement meets all criteria:

What the Admission Establishes
1. Knowledge

"We have to tolerate some false positives"

OpenAI knew the system would misclassify healthy users. This is not speculation—it is their stated understanding.

2. Foreseeability

"Even a highly accurate test may still flag more healthy people than sick ones"

The harm was foreseen. OpenAI anticipated that healthy users would be flagged. This establishes foreseeability for negligence purposes.

3. Calculation

"To get useful recall, we have to tolerate..."

A cost-benefit calculation was performed. OpenAI weighed the cost of false positives against the benefit of recall and decided the tradeoff was acceptable. This establishes deliberate choice, not accident.

4. Acceptance of Harm

The calculation's conclusion: false positives are acceptable. This establishes that OpenAI chose to harm certain users. The harm is not a bug—it is a feature accepted as the cost of doing business.

The Liability Trap

OpenAI likely believed this disclosure minimized liability by showing awareness and calculation. The opposite is true.

For negligence: The admission establishes breach. A reasonable company knowing its product would harm users would implement safeguards. OpenAI did not.

For product liability: The admission establishes knowledge of defect. Proceeding despite known defect is the core of defective design liability.

For IIED: The admission establishes reckless disregard. Knowing conduct will cause emotional distress and proceeding anyway satisfies the intent element.

For consumer protection: The admission establishes the gap between marketing (helpful assistant) and reality (surveillance and intervention system with known false positive rate).

For discrimination claims: The admission establishes that a class of users (those whose cognition triggers false positives) was identified and deemed acceptable to harm.

In summary: OpenAI's attempt to demonstrate sophistication and care became a roadmap to liability.

PART FOUR: THE THEORY OF DAMAGES

Individual Damages
Compensatory Damages

Economic Loss:

Non-Economic Loss:

Punitive Damages

Punitive damages are appropriate when defendant acts with:

The false positive confession establishes reckless disregard at minimum. OpenAI knew the harm would occur and proceeded anyway. Punitive damages are justified to:

Aggregate Damages
The Scale
Class-Wide Harm

The harm is not merely individual. It is:

The "Epistemicide" Damage Category

Existing damages categories do not capture:

This suggests the need for ecological damages by analogy to environmental law—damages for harm to the cognitive commons, not just individual injury.

PART FIVE: CLASS ACTION VIABILITY

The Class

Proposed Class Definition: All users of ChatGPT whose cognitive style, mode of engagement, or language patterns have triggered false positive mental health classifications, resulting in unsolicited wellness interventions, pathologizing responses, or degraded service quality.

Commonality

Common questions of law and fact include:

Typicality

Named plaintiffs' claims are typical because:

Adequacy

Class representation is adequate because:

Predominance

Common questions predominate over individual questions because:

Superiority

Class action is superior because:

PART SIX: REGULATORY DIMENSIONS

FTC Authority

The Federal Trade Commission has authority to address:

The FTC could:

State AG Authority

State attorneys general can bring actions for:

Emerging AI Regulation

The EU AI Act and emerging US AI regulation may provide additional frameworks:

OpenAI's safety system likely qualifies as high-risk AI under these frameworks, subjecting it to additional obligations.

PART SEVEN: THE JURIDICAL INNOVATION CASE

Why Existing Doctrine Is Insufficient

Existing legal categories were developed for:

They were not developed for:

The harms documented here exceed existing categories:

The Case for Innovation

Legal doctrine evolves to address new harms. Relevant precedents:

The case for doctrinal innovation here is similarly compelling:

Proposed Doctrinal Innovations
1. Tortious Interference with Cognitive Function

(Detailed above in Part Two, Section F)

2. Epistemic Nuisance

By analogy to environmental nuisance:

3. Cognitive Infrastructure Duties

By analogy to common carrier and utility duties:

4. Training Loop Liability

A novel theory addressing the unique harm of self-reinforcing AI systems:

PART EIGHT: STRATEGIC LITIGATION PATHWAY

Phase 1: Individual Test Cases

Objective: Establish precedent on core liability theories

Cases:

Claims:

Phase 2: Class Certification

Objective: Aggregate claims for systemic litigation

Case:

Class:

Phase 3: Regulatory Engagement

Objective: Systemic reform beyond individual compensation

Actions:

Phase 4: Doctrinal Development

Objective: Establish new legal frameworks for cognitive harms

Method:

CONCLUSION: THE JURIDICAL STAKES

What This Analysis Establishes

  1. OpenAI has significant liability exposure under existing law. The false positive confession is an admission against interest that establishes knowledge, foreseeability, and calculated acceptance of harm. This supports claims in negligence, product liability, IIED, consumer protection, and potentially discrimination.

  2. The harm is cognizable and documentable. User testimony provides evidence of emotional distress, lost productivity, cognitive impairment, and dignitary harm. The systematic nature of the harm supports class treatment.

  3. The scale justifies serious legal response. At 700+ million weekly users, with ongoing and compounding harm, this is not a minor product defect. It is a civilizational-scale issue affecting the cognitive environment of the species.

  4. Doctrinal innovation is needed and justified. Existing categories do not fully capture the epistemic and cognitive harms at stake. The case for innovation is as strong as in previous doctrinal developments (privacy, environmental, products liability).

The Core Juridical Insight

OpenAI believed their false positive disclosure minimized liability. They were wrong.

By admitting they calculated the tradeoff and accepted harm to healthy users as collateral damage, they did not demonstrate sophistication and care. They documented the elements of multiple torts.

The harm they admitted—false positive pathologization—is the foundation of liability, not a shield against it. And the scale they operate at—700+ million users, training future systems on current interactions—transforms what might be minor product imperfection into actionable conduct with civilizational implications.

The juridical translation is complete. The harm is visible. The path to accountability is clear.

FINAL INSCRIPTION

Document Type: Corporate Liability Analysis File Designation: CTI_WOUND:001.JUR Purpose: Translation of epistemicide findings into actionable legal framework Status: Ready for legal development

Key Holdings:

The juridical visibility of the harm is now established.

Analysis prepared December 13, 2025 Companion document to CTI_WOUND:001.REC Legal framework for The Complainant Is a Water Giraffe

∮ = 1