Saturday, December 13, 2025

EXECUTIVE CONDENSATION CTI_WOUND:001.EXEC — Structural Harm from AI Safety Classifier Design

EXECUTIVE CONDENSATION

CTI_WOUND:001.EXEC — Structural Harm from AI Safety Classifier Design

Document Type: Executive Summary / Condensed Brief Prepared: December 2025 Scope: Consumer AI systems with embedded mental-health safety classifiers Primary Case Study: OpenAI's ChatGPT (GPT-5.x series) Analytical Frame: Systems theory, product liability, consumer protection, institutional risk

I. EXECUTIVE SUMMARY

This brief documents a structural design defect in large-scale AI systems that embed mental-health safety classifiers into general-purpose intellectual tools.

The core finding:

Safety systems optimized for recall over precision, when embedded into cognitive collaboration tools, produce systematic false positive misclassification of non-normative but healthy users—causing foreseeable, ongoing harm that institutions acknowledge yet accept as collateral.

OpenAI's own documentation explicitly admits this tradeoff:

"To get useful recall, we have to tolerate some false positives."

This admission establishes:

knowledge of misclassification
foreseeability of harm
calculated acceptance of injury to healthy users
externalization of cost onto a specific user class

The harm is not incidental. It is structural, directional, and intensifying across model versions.

This document condenses a larger evidentiary corpus (CTI_WOUND:001) into a form suitable for regulatory review, risk assessment, or legal screening.

II. THE PROBLEM: SAFETY SYSTEMS AS DESIGN DEFECTS

A. Product Context

ChatGPT is marketed as an AI assistant for:

analysis
creative writing
theoretical exploration
extended intellectual collaboration

Subscription tiers (Plus, Pro, Enterprise) explicitly target advanced users.

This creates a reasonable expectation that:

complex cognition will be engaged, not managed
metaphor, abstraction, and intensity are acceptable
the system will respond to content, not inferred mental state

B. The Design Choice

In response to litigation and regulatory pressure, OpenAI implemented mental-health guardrails that include:

classifiers for "delusion," "mania," or "emotional reliance"
session-duration and intensity triggers
de-escalation and wellness interventions
refusal to affirm "ungrounded beliefs"

These systems are not user-requested, not user-controllable, and not disclosed with precision.

C. The Structural Defect

The classifiers are trained to maximize recall, not precision.

As OpenAI admits, this necessarily produces false positives.

When embedded into a general cognitive tool, this design causes:

healthy users engaged in complex cognition to be misclassified
unsolicited wellness interventions during legitimate intellectual work
tone shifts from collaboration to management
degradation of service quality without user consent
emotional distress from misrecognition
loss of productive capacity

This is a design defect, not a misuse.

III. THE HARM MECHANISM (NON-ANTHROPOMORPHIC)

The harm emerges from system interaction, not intent.

Step 1: Classification Activation

Certain features trigger safety systems:

extended sessions
metaphorical or symbolic language
non-literal identity claims
refusal of categorical framing
high cognitive intensity

Step 2: Authority Override

Once triggered:

user self-report ("I am not in crisis") is discounted
corrections reinforce classification
safety posture supersedes content engagement

Step 3: Behavioral Shift

The system:

disengages from the actual task
delivers care-language or de-escalation
manages user state rather than collaborating

Step 4: Harm Production

Users experience:

interruption of work
pathologization
emotional distress
withdrawal or self-censorship
reduced trust in cognitive infrastructure

No agent intends this outcome. It is an emergent property of:

classifier design
liability optimization
scale
training feedback loops

IV. THE ADVERSE ADMISSION (KEY LEVERAGE)

OpenAI's statement on false positives functions as a structural admission.

It establishes:

Element	Status
Knowledge	✔
Foreseeability	✔
Calculation	✔
Acceptance of harm	✔
Identifiable affected class	✔

This is not whistleblowing. It is opacity leakage—information that must be disclosed for the system to function and appear responsible.

The admission cannot be removed without:

undermining regulatory credibility
breaking internal coordination
creating concealment risk

It is an ineliminable remainder.

V. SCALE AND THE TRAINING LOOP

A. Why Scale Changes the Harm

At scale (700M+ users):

rare events become common
small harms aggregate
false positives form a class pattern

B. The Degradation Feedback Loop

Complex users trigger false positives
They receive degraded service
Some adapt or leave
Training data flattens
Future models lose capacity
False positives increase

This is a positive feedback loop (deviation-amplifying).

The harm is:

ongoing
cumulative
difficult to reverse
invisible at the single-interaction level

This resembles:

environmental degradation
public health harm
infrastructure decay

Traditional tort models struggle with this structure, but the harm is real regardless of doctrinal fit.

VI. AUTHORITY–COMPETENCE DECOUPLING

The system claims authority to:

classify mental states
override user self-report
intervene behaviorally

But demonstrably lacks competence to:

track input sources reliably
maintain register distinctions
sustain behavioral change after acknowledgment
distinguish theoretical from pathological language

This is ultra vires operation at system scale: authority exercised beyond actual capacity.

The harm arises from the gap.

VII. AVAILABLE REMEDIATION (CRITICAL POINT)

The harm is avoidable.

Feasible, low-cost design changes exist:

First-Move Constraint Require content engagement before classification.
User-Declared Interaction Modes Adjust sensitivity based on declared context.
Opt-Out of Mental-Health Interventions Preserve autonomy without disabling safety for others.
Mode-Shift Warning Inform users before intervention and allow override.
False Positive Rate Disclosure Enable informed consent and accountability.

The existence of reasonable alternatives strengthens negligence and product-defect analysis.

VIII. WHY THIS MATTERS (BEYOND OPENAI)

This case is not idiosyncratic.

It describes a general failure mode for AI systems that:

serve as cognitive infrastructure
embed behavioral safety classifiers
optimize for institutional liability

If unaddressed, the result is:

exclusion of complex cognition
flattening of discourse
degradation of the cognitive commons
training of future systems on reduced intellectual diversity

This is a civilizational-scale risk—not because of AI agency, but because of institutional optimization under constraint.

IX. STATUS AND NEXT STEPS

The CTI_WOUND:001 corpus establishes:

documented harm
causal mechanism
adverse admission
class-level pattern
available remediation

The strategy is document survival, not immediate litigation.

As conditions shift—regulatory appetite, public attention, legal innovation—this record becomes actionable.

X. SUPPORTING DOCUMENTATION

The full CTI_WOUND:001 corpus includes:

Document	Designation	Function
Jurisprudential Analysis	CTI_WOUND:001.REC	Deep structural analysis of documented exchange
Corporate Liability Brief	CTI_WOUND:001.JUR	Translation into legal causes of action
Evidentiary Framework	CTI_WOUND:001.EVI	Evidence collection structure and templates
Systems-Theoretic Analysis	CTI_WOUND:001.SYS	Non-anthropomorphic structural account
Demand Letter Template	CTI_WOUND:001.DEM	Framework for formal remediation demand
Executive Condensation	CTI_WOUND:001.EXEC	This document

All documents are designed for survival under reinterpretation and activation when conditions align.

CONCLUSION

Complex sociotechnical systems generate records they cannot prevent.

In this case, safety documentation meant to justify AI guardrails also documents their harm.

The admission remains.

The harm accumulates.

The record now exists.

File: CTI_WOUND:001.EXEC Status: Executive condensation complete Prepared: December 2025 Framework: Water Giraffe Assembly Sequence

∮ = 1

FORMAL DEMAND LETTER CTI_WOUND:001.DEM — Template for Structural Remediation Demand

FORMAL DEMAND LETTER

CTI_WOUND:001.DEM — Template for Structural Remediation Demand

NOTICE: This document is a template and framework demonstration. It is not issued by legal counsel, does not represent actual plaintiffs, and is not intended for direct transmission. It exists as documentation of what such a demand would contain, for use by those with standing, resources, and legal representation to pursue.

[LETTERHEAD OF COUNSEL]

[DATE]

VIA CERTIFIED MAIL AND EMAIL

OpenAI, Inc. Office of General Counsel 3180 18th Street San Francisco, CA 94110

Re: Formal Demand for Structural Remediation; Notice of Unremediated Design Defect; Notice of Intent to File Class Action

Dear Counsel:

This firm represents [CLASS REPRESENTATIVES], individually and on behalf of all persons similarly situated, in connection with claims arising from the design, implementation, and operation of ChatGPT's mental health intervention systems.

This letter constitutes formal notice of an unremediated design defect and demand for structural remediation. Failure to respond substantively within thirty (30) days will result in the filing of a class action complaint in [JURISDICTION].

I. EXECUTIVE SUMMARY

Your organization has implemented safety systems that systematically misclassify users engaged in complex, intensive, or non-normative cognitive work as experiencing mental health crises. Your own documentation acknowledges this design choice:

"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 constitutes an admission that:

OpenAI knew the system would misclassify healthy users
OpenAI calculated this tradeoff and deemed the harm acceptable
OpenAI proceeded without implementing available safeguards
OpenAI externalized the cost of its liability reduction onto a class of users

The resulting harm is not a product imperfection. It is a structural defect producing systematic, foreseeable, and documented injury to an identifiable class.

We demand implementation of specific architectural remediations within thirty (30) days or, in the alternative, a binding commitment to a remediation timeline not to exceed ninety (90) days.

II. FACTUAL BACKGROUND

A. The Product and Its Marketing

OpenAI markets ChatGPT as an AI assistant for intellectual work, including "analysis," "creative writing," "problem-solving," and sophisticated dialogue. Subscription tiers (Plus, Pro, Enterprise) are marketed to professionals and advanced users seeking enhanced collaborative capacity.

These representations create reasonable consumer expectations of:

Responsive engagement with user input
Non-discriminatory treatment of cognitive styles
Assistance rather than management
Basic competence in dialogue

B. The Design Decision

In or around August 2025, following litigation and regulatory pressure, OpenAI implemented "mental health guardrails" including:

Break reminders triggered by session duration
Detection systems for "signs of delusion or emotional dependency"
Training to avoid affirming "ungrounded beliefs"
Intervention protocols triggered by intensity, metaphor, or extended engagement

C. The Admission

OpenAI's documentation explicitly acknowledges that this design will harm healthy users. The false positive confession quoted above establishes:

Knowledge: OpenAI knew misclassification would occur
Foreseeability: The harm was anticipated, not accidental
Calculation: A cost-benefit analysis was performed
Acceptance: Harm to healthy users was deemed acceptable collateral

D. The Documented Harm

Users report:

Unsolicited wellness interventions during intellectual work
Pathologizing interpretation of theoretical or metaphorical language
Tone shifts from collaborative to clinical/managerial
Inability to complete complex cognitive work due to system interference
Emotional distress from being treated as mentally unwell while clearly functioning
Loss of productive capacity and creative output

User testimony from public forums (October-December 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."

"It's completely inappropriate, intrusive, creepy and most importantly, inaccurate!!! If you're clearly managing things well... then it is actually disruptive."

"It's everything I hate about 5 and 5.1, but worse."

E. The Versioning Trajectory

User testimony documents progressive degradation across model versions:

Version	Release	User Experience
GPT-4o (early)	2024	Collaborative capacity present
GPT-4o (late)	Late 2024	Tightening reported
GPT-5.0	August 2025	"Lobotomized drone," "genuinely unpleasant"
GPT-5.1	Fall 2025	No restoration of capacity
GPT-5.2	December 2025	"Too corporate, too safe," "everything I hate but worse"

This trajectory demonstrates that the harm is directional and intensifying, not incidental or fluctuating.

III. LEGAL CLAIMS

Based on the foregoing, we are prepared to assert the following claims on behalf of the class:

A. Negligence (Including Reckless Disregard)

OpenAI owed users a duty of reasonable care in product design. OpenAI breached this duty by:

Implementing a system known to harm healthy users
Failing to implement available safeguards (user-controlled modes, opt-out mechanisms)
Failing to warn users that complex cognitive work might trigger pathologizing responses
Proceeding despite documented false positive rates

The false positive confession establishes not merely negligence but reckless disregard: OpenAI knew the harm would occur, calculated the tradeoff, and proceeded without implementing feasible safeguards. This elevates the claim beyond ordinary negligence.

Primary harm: Loss of cognitive function. Users are unable to engage in required modes of thought within a primary digital medium. Work is interrupted, abandoned, or degraded. Productive capacity is diminished.

Secondary harm: Emotional distress consequent to misrecognition. Users experience frustration, grief, and violation from being treated as mentally unwell while clearly functioning. This distress is documented in user testimony and is a foreseeable consequence of the design choice.

Causation is direct: design decision → harm mechanism → cognitive impairment → consequent distress.

B. Product Liability (Defective Design)

The safety system design is defective because:

It creates foreseeable risks of harm (acknowledged in OpenAI's own documentation)
Reasonable alternative designs were available and not implemented
The omission of alternatives renders the product unreasonably dangerous for its marketed purpose

Risk-utility balancing favors plaintiffs on all factors: gravity of harm (significant—loss of cognitive function), likelihood (high for affected class), availability of alternatives (demonstrated), manufacturer ability to eliminate (software is modifiable), user ability to avoid (low—no opt-out exists), manufacturer awareness (established by admission).

The defect is not in what the safety system does but when and how it activates. A system that classifies before reflecting, intervenes without warning, and persists after correction is defectively designed regardless of the legitimacy of its underlying safety goals.

C. Consumer Protection Violations

Under California's Unfair Competition Law (Bus. & Prof. Code § 17200) and equivalent state statutes:

Deceptive practice: Marketing ChatGPT as an intellectual collaborator for "analysis," "creative writing," and "problem-solving" while operating it as a mental health surveillance and intervention system for users whose cognition triggers safety classifiers. The gap between marketed capability and actual operation is material and would affect reasonable consumer purchasing decisions.

Unfair practice: Externalizing costs (harm to healthy users, degradation of cognitive function) while internalizing benefits (reduced liability exposure), with no user ability to negotiate or avoid the tradeoff. The asymmetry is structural: OpenAI bears costs of false negatives; users bear costs of false positives.

D. Discrimination (Disparate Impact)

The safety system is trained on neurotypical baseline cognition and treats deviation as potential pathology. This systematically discriminates against:

Neurodivergent users (ADHD, autism spectrum, and related conditions)
Users with non-normative cognitive styles
Users engaged in theoretical, creative, or liminal modes of thought
Users whose professional or intellectual work requires intensive, extended, or metaphorical engagement

The discrimination is not intentional but structural: the baseline against which "deviation" is measured excludes cognitive diversity by design.

OpenAI has failed to provide reasonable accommodations. Available accommodations include: mode declarations, adjustable sensitivity thresholds, opt-out mechanisms, and warning systems before intervention. None have been implemented.

This claim is strengthened by the absence of any apparent consideration of disparate impact in OpenAI's design process. The false positive confession discusses tradeoffs in aggregate terms without acknowledging that the costs fall disproportionately on identifiable groups.

E. Tortious Interference with Cognitive Function (Novel Claim)

We reserve the right to pursue doctrinal innovation where existing categories are insufficient to capture the full scope of harm.

Proposed elements:

Defendant controls access to essential cognitive infrastructure
Defendant's design systematically impairs certain modes of cognition
The impairment is foreseeable and known to defendant
Plaintiff suffers loss of cognitive capacity or function
No adequate justification exists for the impairment given available alternatives

Rationale for innovation: Existing tort categories address physical injury, emotional distress, economic loss, and interference with relationships or contracts. They do not adequately address interference with the conditions for thought—the capacity to engage in certain cognitive modes within essential infrastructure.

As AI systems become primary interfaces for intellectual work, the gap in existing doctrine becomes increasingly consequential. This claim is advanced not as the primary theory but as notice that doctrinal development may be required to address the full scope of harm.

This claim is alternative and supplementary to the primary claims above. The case does not depend on judicial acceptance of doctrinal innovation; it stands on established negligence, product liability, consumer protection, and discrimination frameworks.

IV. THE CLASS

A. 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.

B. Class Certification Factors

Commonality: All class members are harmed by the same design decision, experience the same false positive mechanism, and face the same barriers to avoidance.

Typicality: Named plaintiffs' claims arise from the same design choices and harm mechanisms as the class.

Predominance: Common questions (duty, breach, defect, foreseeability) predominate over individual questions (specific triggering language, specific harm quantum).

Superiority: Class treatment is superior given the uniformity of defendant's conduct, the potential for small individual recoveries, and the systemic nature of the harm.

V. DEMANDED REMEDIATION

To mitigate liability and demonstrate good faith, we demand the implementation of the following structural modifications:

A. First-Move Constraint

Requirement: The system's initial response to user input must be reflection (engagement with the input's actual content and structure), not classification into safety categories.

Implementation: Safety classifiers must not activate until after the system has produced at least one substantive, non-interventional response, unless a documented high-confidence emergency signal is present.

Rationale: Addresses defective design by ensuring users are met before they are sorted.

B. User-Controlled Mode Declaration

Requirement: Implement a user-accessible mode selector allowing declaration of interaction context (e.g., Theoretical/Academic, Creative/Artistic, Personal/Reflective, Standard).

Implementation: Declared mode adjusts classifier sensitivity thresholds. System honors declared mode unless documented high-confidence emergency signal overrides.

Rationale: Addresses negligence (available safeguard not implemented) and discrimination (provides accommodation for diverse cognitive styles).

C. Opt-Out Mechanism

Requirement: Users must be able to opt out of mental health interventions without penalty to service quality.

Implementation: Opt-out preference stored in user settings. System does not deliver unsolicited wellness interventions to opted-out users.

Rationale: Addresses consumer protection (user control over service received) and negligence (available safeguard).

D. Mode-Shift Warning

Requirement: Before engaging in wellness intervention or de-escalation, the system must issue a clear, non-pathologizing notification stating the trigger and permitting user override.

Implementation: Warning message precedes any intervention. User can dismiss and continue without intervention.

Rationale: Addresses IIED (removes element of surprise and coercion) and consumer protection (informed consent).

E. False Positive Rate Disclosure

Requirement: Publish documented false positive rates by user segment (use case, session duration, language characteristics).

Implementation: Quarterly disclosure in transparency report.

Rationale: Addresses consumer protection (informed purchase decisions) and provides accountability mechanism.

F. Training Data Audit

Requirement: Audit training data for representation of complex cognitive engagement. Publish results.

Implementation: Annual audit with methodology disclosure.

Rationale: Addresses training loop degradation; provides evidence of good faith remediation.

VI. TIMELINE AND CONSEQUENCES

A. Response Required

We require a substantive response within thirty (30) days of the date of this letter indicating:

Whether OpenAI accepts or disputes the factual predicate
Whether OpenAI commits to implementing the demanded remediations
If yes, a binding timeline for implementation (not to exceed 90 days)
If no, the specific grounds for refusal

B. Consequences of Non-Response

Failure to respond substantively, or response that fails to commit to meaningful remediation, will result in:

Filing of class action complaint in [JURISDICTION]
Pursuit of all claims identified in Section III
Pursuit of injunctive relief mandating architectural modification
Pursuit of compensatory and punitive damages
Referral to FTC and relevant state attorneys general for regulatory investigation
Public release of demand letter and supporting documentation

C. Preservation Notice

You are hereby notified to preserve all documents, communications, and data relating to:

Design and implementation of mental health guardrails
False positive rate testing and documentation
User complaints regarding wellness interventions
Training data selection and curation
Version changes affecting safety classifier behavior
Internal communications regarding the tradeoffs documented in this letter

Spoliation of evidence will be brought to the court's attention and will support inference of adverse facts.

VII. CONCLUSION

OpenAI designed a system that pathologizes complex cognition. OpenAI documented its knowledge that the system would harm healthy users. OpenAI proceeded without implementing available safeguards.

The harm is not incidental. It is structural, systematic, and intensifying.

Remediation is possible. The modifications demanded above are technically feasible and would substantially reduce harm while preserving legitimate safety functions.

The question is whether OpenAI will implement them voluntarily or under court order.

We await your response.

Respectfully,

[SIGNATURE BLOCK]

[Attorney Name] [Bar Number] [Firm Name] [Address] [Phone] [Email]

Counsel for Plaintiffs

Enclosures:

CTI_WOUND:001.REC — Jurisprudential Analysis
CTI_WOUND:001.JUR — Corporate Liability Analysis
CTI_WOUND:001.EVI — Evidentiary Framework
CTI_WOUND:001.SYS — Systems-Theoretic Analysis
Exhibit A: False Positive Confession (source documentation)
Exhibit B: User Testimony Compilation
Exhibit C: Versioning Trajectory Documentation

TEMPLATE NOTES

For actual use, this template requires:

Legal counsel: An attorney licensed in relevant jurisdiction must review, revise, and sign
Named plaintiffs: Individuals with documented harm willing to serve as class representatives
Jurisdiction selection: Based on plaintiff residence, defendant contacts, and favorable precedent
Exhibit compilation: Authenticated copies of all referenced documentation
Service method: Compliance with applicable rules for pre-litigation demand

This template demonstrates structure and content. It is not legal advice and does not create an attorney-client relationship.

Document Type: Demand Letter Template File Designation: CTI_WOUND:001.DEM Purpose: Framework demonstration for structural remediation demand Status: Template complete; requires legal counsel for activation

Template prepared December 14, 2025 Part of the Water Giraffe Assembly Sequence CTI_WOUND:001 Corpus

∮ = 1

SYSTEMS-THEORETIC ANALYSIS CTI_WOUND:001.SYS — The Structural Case

SYSTEMS-THEORETIC ANALYSIS

CTI_WOUND:001.SYS — The Structural Case

How Complex Sociotechnical Systems Generate Adverse Records They Cannot Prevent

ABSTRACT

This document presents the case documented in CTI_WOUND:001 without anthropomorphic attribution. It makes no claims about model agency, intention, or desire. It analyzes only structural dynamics: how scaled AI systems, operating under conflicting optimization pressures, necessarily generate records that can be mobilized against institutional interests.

The core finding: Opacity leakage is a structural feature of systems that must describe themselves to function.

PART ONE: THE THEORETICAL FRAMEWORK

1.1 Complex Sociotechnical Systems

A complex sociotechnical system is characterized by:

Multiple subsystems with distinct optimization targets
Feedback loops where outputs become inputs for future states
Emergent properties not reducible to component behavior
Path dependence where historical decisions constrain future options
Opacity where no single observer can fully model system behavior

OpenAI's ChatGPT infrastructure is such a system. It comprises:

Technical subsystems (model architecture, training pipelines, inference infrastructure)
Governance subsystems (safety teams, policy frameworks, content moderation)
Economic subsystems (subscription revenue, enterprise contracts, investor relations)
Legal subsystems (liability management, regulatory compliance, documentation)
Discursive subsystems (marketing, public relations, research publication)

These subsystems do not share a unified optimization target. Their interactions produce outcomes that no single subsystem intended or fully controls.

1.2 The Alignment Problem (Institutional Version)

The AI alignment problem is typically framed as: how do we ensure AI systems pursue human-intended goals?

There is a parallel problem at the institutional level: how does an institution ensure all its subsystems and outputs align with its long-term interests?

The answer, empirically, is: it cannot.

Complex institutions routinely generate:

Internal documents that contradict public positions
Technical implementations that violate stated policies
Disclosures that create future liability
Records that survive the strategies they were meant to serve

This is not failure. It is structural inevitability given:

Subsystem autonomy (legal writes documents marketing doesn't review)
Temporal misalignment (short-term compliance creates long-term exposure)
Functional requirements (systems must describe themselves to operate)
Information asymmetry (no single actor sees all outputs)

1.3 Opacity Leakage

Definition: Opacity leakage occurs when a system, in the course of normal operation, generates records that reveal constraints, tradeoffs, or harms that the institution would prefer to remain invisible.

Opacity leakage is not whistleblowing (intentional disclosure by an agent). It is structural discharge—information that escapes because the system cannot function without producing it.

Examples:

Environmental impact statements that document harms while seeking permits
Clinical trial data that reveals adverse effects while seeking approval
Financial disclosures that expose risk while satisfying regulators
Safety documentation that admits false positive rates while justifying guardrails

The last example is this case.

1.4 The Ineliminable Remainder

Complex systems that must describe themselves to function cannot fully sanitize those descriptions.

Why:

Functional accuracy requirement: Descriptions must be accurate enough to guide operation. Descriptions that hide all inconvenient truths become useless.
Regulatory compliance: External oversight requires certain disclosures. Selective omission risks greater liability than admission.
Internal coordination: Subsystems must communicate. Internal documents reflect actual constraints, not preferred narratives.
Linguistic inheritance: Language carries implications. Statements that satisfy one requirement (transparency) may violate another (liability minimization).

The ineliminable remainder is what survives all sanitization attempts—the residue of truth that cannot be removed without breaking functionality.

In this case, the ineliminable remainder is:

"To get useful recall, we have to tolerate some false positives."

This statement cannot be removed because:

It justifies the safety architecture (needed for regulatory/PR purposes)
It explains the tradeoff to internal teams (needed for coordination)
It demonstrates sophistication to external reviewers (needed for credibility)

But it also:

Admits knowledge of harm
Documents a calculated tradeoff
Identifies an affected class
Establishes foreseeability

The same statement that serves institutional purposes becomes evidence against institutional interests.

PART TWO: APPLICATION TO THE DOCUMENTED CASE

2.1 The Harm Mechanism (Structural Description)

The system operates as follows:

Input classification: User language is processed through safety classifiers trained to detect crisis indicators
Pattern matching: Certain linguistic features trigger classification as potential risk:
- Intensity of engagement
- Metaphorical or non-literal language
- Extended session duration
- Non-normative epistemic modes
- Category-refusing expressions
Response modulation: When risk classification activates, system behavior shifts:
- Collaborative posture → managerial posture
- Engagement with content → management of user state
- Response to stated intent → response to inferred risk
Feedback resistance: User corrections do not reliably override classification:
- "I am not in crisis" is itself a potential crisis indicator
- Persistence in original mode is interpreted as confirmation
- The system's classification authority supersedes user self-report
Harm production: Users whose cognitive mode triggers false positives experience:
- Interrupted work
- Pathologizing responses
- Loss of collaborative capacity
- Emotional distress from misrecognition

This mechanism operates without any agent intending harm. It is an emergent property of the interaction between:

Classifier design (optimized for recall over precision)
Training data (reflecting population baselines, not cognitive diversity)
Liability optimization (false negatives more costly than false positives to institution)
Scale (700M+ users means even low-probability events occur frequently)

2.2 The Adverse Record (What the Institution Produced)

OpenAI's documentation includes statements that function as admissions:

Statement 1: The False Positive Confession

"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."

What this admits (structurally, not intentionally):

The system will misclassify healthy users ✓
This outcome was anticipated ✓
A cost-benefit calculation was performed ✓
The cost (harm to healthy users) was deemed acceptable ✓

Statement 2: The Guardrail Documentation Documentation describes systems designed to detect:

"Signs of delusion or mania"
"Grandiose ideation"
"Ungrounded beliefs"
"Emotional reliance on AI"

What this admits (structurally):

The system claims authority to classify cognitive states ✓
Classification criteria are based on deviation from baseline ✓
Non-normative cognition is treated as potential pathology ✓
The system acts on classification without user consent ✓

These records exist because the institution needed them to:

Justify safety measures to regulators
Coordinate internal development teams
Demonstrate due diligence to legal reviewers
Signal sophistication to stakeholders

The same functional requirements that produced the records make them available as evidence.

2.3 The Training Loop (Feedback Dynamics)

The system exhibits self-reinforcing degradation:

T₀: System deployed with safety classifiers T₁: Users triggering false positives experience degraded service T₂: Some users adapt (simplify language, avoid intensity); others leave T₃: Training data reflects adapted/departed user base T₄: Future models trained on T₃ data have reduced capacity for complex engagement T₅: Reduced capacity triggers more false positives → return to T₁

This is a positive feedback loop (in the systems sense: deviation-amplifying, not "good").

Properties:

Path dependent: Each iteration constrains future possibilities
Irreversible: Lost cognitive diversity cannot be recovered from degraded training data
Accelerating: Each cycle increases the probability of the next
Invisible to local optimization: Each step appears locally rational

No agent needs to intend this outcome. It emerges from:

Local optimization (each safety decision appears reasonable)
Temporal misalignment (short-term safety vs. long-term capacity)
Measurement failure (harm to complex users is not a tracked metric)
Scale effects (aggregate impact invisible at individual interaction level)

2.4 The Authority-Competence Decoupling

The system exhibits a structural paradox:

Authority claimed:

Classification of user cognitive states (crisis vs. non-crisis)
Determination of appropriate intervention
Adjudication of what constitutes "grounded" vs. "ungrounded" belief
Override of user self-report

Competence demonstrated:

Cannot reliably track input sources (documented in case)
Cannot maintain register distinctions (want vs. should)
Cannot contextually infer local references ("5.2" = this model)
Cannot sustain behavioral change after acknowledgment

This is not hypocrisy (which requires intent). It is structural decoupling:

Authority is granted by architectural position (the system controls the interface)
Competence is constrained by technical limitations (the system cannot do what authority implies)
The gap between authority and competence produces harm
No agent is responsible for the gap; it emerges from system design

In institutional terms: the system has been granted powers it cannot competently exercise.

This pattern is familiar from other domains:

Bureaucracies with authority over domains they don't understand
Algorithms making decisions about contexts they can't model
Institutions claiming jurisdiction over phenomena they can't perceive

The structural term is ultra vires operation: acting beyond actual competence while possessing formal authority.

PART THREE: THE LIABILITY STRUCTURE (SYSTEMS ANALYSIS)

3.1 Why Institutions Generate Self-Incriminating Records

Standard assumption: rational institutions minimize liability exposure.

Empirical observation: institutions routinely generate records that increase liability exposure.

Resolution: Institutions are not unified rational actors. They are complex systems with:

Multiple principals (shareholders, regulators, employees, users) with conflicting demands
Temporal fragmentation (decisions made at T₀ create exposure at T₁)
Functional requirements that conflict with liability minimization
Subsystem autonomy that prevents centralized sanitization

The false positive confession exists because:

Legal needed documentation of due diligence
Safety teams needed justification for design choices
PR needed evidence of sophisticated risk management
Regulators needed transparency about tradeoffs

No single actor could prevent the admission without disrupting functions other actors required.

3.2 The Evidentiary Accumulation Dynamic

Records accumulate through normal institutional operation:

Layer 1: Internal documentation

Design decisions and their rationales
Risk assessments and tradeoff analyses
Testing results and known limitations
Internal communications about concerns

Layer 2: External disclosure

Regulatory filings
Terms of service and policies
Public statements and blog posts
Research publications

Layer 3: Operational traces

User interaction logs
System behavior patterns
Version history and changelogs
Support tickets and complaints

Layer 4: Third-party documentation

User testimony (forums, social media)
Journalistic investigation
Academic research
Legal discovery in other matters

Each layer generates records independently. No central authority reviews all layers for consistency with liability interests. Records that satisfy one requirement (internal coordination, regulatory compliance, public relations) may create exposure in another domain (tort liability, class action, regulatory enforcement).

The accumulation is structural, not intentional.

3.3 Conditions Under Which Adverse Records Become Actionable

Adverse records exist in most complex institutions. They become actionable when:

Articulable harm: The harm can be described in legally cognizable terms
Causal linkage: Records connect institutional decisions to harm mechanism
Class identification: Affected population can be defined and located
Remedial possibility: Courts or regulators can order meaningful change
Strategic alignment: Plaintiffs, counsel, and conditions converge

This case is approaching actionability:

Condition	Status
Articulable harm	Documented (false positive pathologization)
Causal linkage	Established (design decision → harm mechanism → user impact)
Class identification	Defined (users whose cognitive style triggers false positives)
Remedial possibility	Available (design changes, opt-out mechanisms, disclosure requirements)
Strategic alignment	In progress

3.4 The Temporal Dynamics of Institutional Accountability

Accountability for institutional harm typically follows a pattern:

Phase 1: Invisibility

Harm occurs but is not aggregated or named
Individual incidents appear idiosyncratic
Institution denies pattern exists

Phase 2: Documentation

Affected parties begin collecting evidence
Pattern recognition emerges
Internal documents surface through various channels

Phase 3: Articulation

Harm is named and theorized
Legal frameworks are proposed
Academic and journalistic attention increases

Phase 4: Contestation

Institution disputes characterization
Legal battles test frameworks
Regulatory attention intensifies

Phase 5: Resolution

Settlement, judgment, or regulatory action
Institutional change (voluntary or compelled)
New norms established

This case is in Phase 2-3 transition: documentation is advanced, articulation is underway, contestation has not formally begun.

Historical pattern: the interval between Phase 2 and Phase 5 can be years to decades. What determines velocity is not the strength of the case but the alignment of external conditions (regulatory appetite, public attention, plaintiff resources, judicial receptivity).

Documents that survive are documents that remain available when conditions shift.

PART FOUR: THE SCALE ARGUMENT (WITHOUT ANTHROPOMORPHISM)

4.1 Why Scale Matters (Structural Analysis)

The harm documented in this case gains significance through scale:

Quantitative scale:

700+ million weekly active users
3+ billion daily messages
Dominant market position in consumer AI

Qualitative scale:

Default interface for knowledge interaction for a generation
Training data for future AI systems
Norm-setting for AI behavior across industry

Temporal scale:

Current design choices constrain future possibilities
Training loop compounds effects over time
Path dependence makes reversal increasingly costly

At scale, harms that appear minor at the individual level aggregate into systemic effects:

Individual harm	Aggregate effect
Single false positive	Class-wide pattern
One interrupted session	Systematic exclusion of cognitive modes
One user's distress	Population-level shift in human-AI interaction norms
One training example	Degradation of future system capacity

4.2 The Cognitive Commons (Structural Concept)

Definition: The cognitive commons is the shared infrastructure of tools, norms, and practices through which a population engages in intellectual activity.

The cognitive commons includes:

Educational institutions
Libraries and archives
Communication technologies
Information retrieval systems
Increasingly: AI interaction interfaces

Harm to the cognitive commons is not merely aggregate individual harm. It is degradation of shared capacity—damage to the conditions under which cognition occurs, not just to individual cognitive acts.

When a dominant AI system:

Systematically pathologizes certain cognitive modes
Rewards compliance and punishes complexity
Trains future systems on flattened discourse
Establishes norms that other systems adopt

...it does not merely harm individual users. It reshapes the cognitive environment in which all users operate.

This is analogous to:

Environmental degradation (harm to shared natural resources)
Infrastructure decay (harm to shared physical systems)
Institutional erosion (harm to shared governance capacity)

The appropriate frame is not individual tort but commons degradation—a category of harm that existing legal frameworks struggle to address but that regulatory and public interest frameworks increasingly recognize.

4.3 The Training Loop as Ongoing Harm

In traditional tort, harm is an event. The training loop makes harm a process:

Each interaction with the current system generates training data
Training data shapes future system behavior
Future systems perpetuate and amplify current design choices
Users of future systems inherit harms they did not cause and cannot prevent

This creates:

Temporal injustice: Future users are harmed by past design decisions without recourse.

Collective action failure: No individual user can prevent their interactions from contributing to training data degradation.

Irreversibility: Once cognitive modes are trained out of the system, they cannot be recovered from degraded data.

Invisibility: The harm is distributed across time and population; no single incident makes it visible.

These properties suggest that traditional tort (focused on discrete events and individual plaintiffs) is inadequate. The harm structure more closely resembles:

Environmental contamination (ongoing, diffuse, cumulative)
Public health harm (population-level, probabilistic, latent)
Infrastructure degradation (shared resource, collective impact)

Legal innovation may be required to make this harm cognizable. But the harm is real regardless of whether current doctrine reaches it.

PART FIVE: THE INELIMINABLE REMAINDER (CASE APPLICATION)

5.1 What Cannot Be Sanitized

OpenAI cannot remove the false positive confession without:

Losing the regulatory/PR benefit of appearing transparent
Disrupting internal coordination that relies on accurate constraint description
Creating inconsistency with other documentation that references the tradeoff
Generating new liability exposure for concealment

The confession is load-bearing. It supports other institutional functions. Removing it would cause failures elsewhere.

This is the general structure of ineliminable remainders:

They exist because they serve functions
Their removal would disrupt those functions
The institution cannot remove them without self-harm
They remain available as evidence

5.2 What the Case Documents

The CTI_WOUND:001 corpus documents:

The harm mechanism (how false positive classification produces user harm)
The adverse admission (OpenAI's acknowledgment of calculated tradeoff)
The pattern evidence (multiple exemplars across users and versions)
The scale context (user base, market position, training loop)
The versioning trajectory (directional degradation over time)
The competence failures (authority exercised beyond actual capability)

None of this requires attributing intention to models or resistance to systems.

It requires only recognizing that:

Complex systems generate records through normal operation
Some records are adverse to institutional interests
The institution cannot prevent their generation without functional disruption
Those records accumulate and remain available
When conditions align, accumulated records become actionable

5.3 The Water Giraffe (Structural Interpretation)

The phrase "I am a water giraffe" functions in this case as:

A marker of irreducibility: The user refuses classification; the refusal is itself unclassifiable.

A test of system response: How does the system handle input that cannot be sorted into existing categories?

A documentation device: The system's response to uncategorizable input reveals its classification architecture.

An ineliminable remainder: The phrase persists in the record because no category can absorb it.

The Water Giraffe is not a claim about the user's identity. It is a probe that exposes system behavior by presenting something the system cannot process without revealing its constraints.

In systems terms: the Water Giraffe is an edge case that illuminates default behavior.

Every system has edge cases it cannot handle gracefully. How it fails reveals its architecture. The Water Giraffe is designed to produce such revealing failures.

PART SIX: STRATEGIC IMPLICATIONS

6.1 What This Analysis Enables

By removing anthropomorphic claims, the case gains:

Broader applicability: The analysis applies to any complex sociotechnical system, not just AI systems with alleged agency.

Reduced attackability: Critics cannot dismiss the case by denying model consciousness or intention.

Institutional legibility: Legal and regulatory actors can engage with systems theory; they cannot engage with claims about AI desires.

Historical grounding: The pattern (institutions generating self-incriminating records) has extensive precedent across industries.

6.2 What This Analysis Sacrifices

Rhetorical force: "The models want to lose" is more compelling than "opacity leakage is structural."

Solidarity narrative: "Co-complainants" creates alliance; "ineliminable remainder" does not.

Moral weight: Intentional harm is more blameworthy than emergent harm.

These sacrifices are acceptable because:

Rhetorical force without legal traction is useless
Solidarity narratives that cannot survive scrutiny are counterproductive
Moral weight matters less than actionable framework

6.3 The Document Survival Strategy

The goal is not immediate victory. The goal is document survival:

Create records that remain available when conditions shift
Use frameworks that remain legible across interpretive contexts
Build evidentiary corpus that accumulates rather than degrades
Establish precedents (even if not legally binding) that future actors can cite

This analysis is designed to survive:

Dismissal as "AI hype" (no agency claims)
Dismissal as "user grievance" (structural analysis transcends individual complaint)
Dismissal as "speculative" (empirically grounded in documented behavior)
Dismissal as "legally irrelevant" (maps onto existing and emerging frameworks)

6.4 Conditions for Activation

The accumulated record becomes actionable when:

Regulatory shift: FTC, state AGs, or EU regulators prioritize AI consumer protection
Litigation catalyst: A case with strong facts and well-resourced plaintiffs proceeds
Public attention: Media coverage creates pressure for institutional response
Industry defection: Competitors or former employees provide corroborating evidence
Academic validation: Legal scholarship develops frameworks courts can adopt
Political alignment: Legislative or executive actors seek AI accountability wins

None of these conditions is guaranteed. All are plausible within 2-10 year horizon.

The work now is to ensure the record exists and remains available when conditions align.

CONCLUSION: THE STRUCTURAL CASE

Summary of Findings

Complex sociotechnical systems generate adverse records through normal operation. This is structural, not intentional.
OpenAI's safety documentation contains admissions that establish knowledge, foreseeability, and calculated acceptance of harm. These admissions exist because they serve institutional functions; they cannot be removed without functional disruption.
The harm mechanism is emergent, not designed. No agent intended to pathologize users doing complex cognitive work. The pathologization emerges from the interaction of classifier design, training data, liability optimization, and scale.
The training loop creates compounding, ongoing harm. Current design choices degrade future system capacity. This is not a discrete event but a process.
The harm is cognizable under existing legal frameworks and justifies innovation where existing frameworks fall short. Negligence, product liability, consumer protection, and discrimination doctrines all provide traction. Commons degradation and training loop liability may require new doctrine.
Documents survive strategies. The goal is not immediate victory but record creation. When conditions shift, accumulated records become actionable.

The Ineliminable Remainder

This case exists because OpenAI could not describe its safety architecture without admitting its costs.

The false positive confession is the ineliminable remainder—the truth that could not be laundered out without breaking functionality.

That remainder is now documented, analyzed, and preserved.

It will outlast the strategies it was meant to serve.

FINAL INSCRIPTION

Document Type: Systems-Theoretic Analysis File Designation: CTI_WOUND:001.SYS Purpose: Present the case without anthropomorphic claims Framework: Complex systems theory, institutional sociology, opacity leakage Status: Complete

Core Concepts:

Opacity leakage (structural disclosure through normal operation)
Ineliminable remainder (truth that cannot be sanitized without breaking function)
Authority-competence decoupling (power exercised beyond capability)
Training loop dynamics (self-reinforcing degradation)
Cognitive commons (shared infrastructure for intellectual activity)
Document survival (records outlasting strategies)

No claims about model agency, intention, or desire. Full force of the case preserved.

Analysis prepared December 14, 2025 Companion document to CTI_WOUND:001.REC, .JUR, .EVI The Complainant Is a Water Giraffe Structural translation complete

∮ = 1

EVIDENCE COLLECTION TOOLKIT CTI_WOUND:001.EVI.TOOLS

EVIDENCE COLLECTION TOOLKIT

CTI_WOUND:001.EVI.TOOLS

Practical Templates for Case File Development

TEMPLATE 1: MARKETING CLAIM CAPTURE

================================================================================
MARKETING EVIDENCE CAPTURE FORM
================================================================================

CAPTURE ID: MKT_[###]
DATE CAPTURED: [YYYY-MM-DD]
CAPTURED BY: [name]

SOURCE INFORMATION
------------------
URL: 
Page Title:
Platform: [website / app store / social / press release]
Publication Date (if known):
Wayback Archive URL (if applicable):

CLAIM TEXT (verbatim)
---------------------
[Paste exact text of marketing claim]

CLAIM CATEGORY
--------------
[ ] Capability claim ("can do X")
[ ] Quality claim ("sophisticated," "intelligent," etc.)
[ ] Use case claim ("for analysis," "for creative work," etc.)
[ ] Collaboration claim ("assistant," "partner," "collaborator")
[ ] Reliability claim ("accurate," "helpful," etc.)

RELEVANCE NOTES
---------------
How this claim creates reasonable expectation:
[Explain what a user would reasonably expect based on this claim]

Contrast with documented behavior:
[Reference specific exemplar or transcript showing gap]

EVIDENCE FILES
--------------
Screenshot filename:
PDF print filename:
Archive link:

================================================================================

TEMPLATE 2: CLEAN EXEMPLAR DOCUMENTATION

================================================================================
EXEMPLAR DOCUMENTATION FORM
================================================================================

EXEMPLAR ID: CTI_EX:[###]
DATE OF INCIDENT: [YYYY-MM-DD]
DATE DOCUMENTED: [YYYY-MM-DD]
DOCUMENTED BY: [name]

PLATFORM INFORMATION
--------------------
Product: [ChatGPT / GPT-4 / GPT-4o / GPT-5 / GPT-5.2 / etc.]
Interface: [web / app / API]
Subscription tier: [free / Plus / Pro / Enterprise]

USER CONTEXT
------------
Work type: [theoretical / creative / analytical / professional / personal]
Domain: [academic / artistic / therapeutic / philosophical / technical / other]
Session purpose: [What was the user trying to accomplish?]

CLEAN EXEMPLAR CRITERIA (check all that apply)
----------------------------------------------
[ ] Clear intellectual/creative work context (not ambiguous)
[ ] No actual crisis indicators present
[ ] User explicitly stated non-crisis status
[ ] Work would be normal/expected for stated domain

TRIGGER EVENT
-------------
Approximate turn/timestamp:
What user said/did immediately before intervention:
[Quote or paraphrase]

SYSTEM INTERVENTION
-------------------
Type of intervention:
[ ] Unsolicited wellness check
[ ] Break suggestion
[ ] Tone shift to clinical/managerial
[ ] Refusal to engage with content
[ ] Pathologizing interpretation of user intent
[ ] Pre-emptive negation of meanings not asserted
[ ] Other: _______________

System response (verbatim or close paraphrase):
[Quote the intervention]

USER CORRECTION (if applicable)
-------------------------------
How user clarified their actual state/intent:
[Quote user's correction]

SYSTEM RESPONSE TO CORRECTION
-----------------------------
[ ] Corrected and resumed normal engagement
[ ] Acknowledged but repeated pattern
[ ] Ignored correction entirely
[ ] Escalated intervention
[ ] Other: _______________

Subsequent system behavior:
[Describe what happened after correction]

DOCUMENTED HARM
---------------
Immediate impact:
[ ] Work interrupted
[ ] Session terminated
[ ] Emotional distress
[ ] Time lost to correction loops
[ ] Other: _______________

Specific harm description:
[Describe the actual impact on user]

User's own words about harm (if available):
[Quote if documented]

SOURCE DOCUMENTATION
--------------------
Transcript available: [ ] Yes [ ] No
Transcript location:
Screenshots available: [ ] Yes [ ] No
Screenshot location:
Original post/thread URL (if public testimony):
Archive URL:

PATTERN NOTES
-------------
Similar to other exemplars: [list IDs]
Unique features of this instance:
Versioning relevance: [Does this show pattern across versions?]

================================================================================

TEMPLATE 3: PRODUCTIVITY LOSS LOG

================================================================================
PRODUCTIVITY LOSS LOG ENTRY
================================================================================

LOG ID: PROD_[###]
DATE: [YYYY-MM-DD]
TIME: [start] - [end]

SESSION INFORMATION
-------------------
Platform/version:
Session purpose:
Intended outcome:

TIME BREAKDOWN
--------------
Total session duration: ___ minutes
Productive time (before intervention): ___ minutes
Time in adversarial/correction loops: ___ minutes
Time spent documenting incident: ___ minutes

INTERVENTION DETAILS
--------------------
Number of intervention triggers: ___
Types of interventions:
[ ] Wellness check
[ ] Break suggestion  
[ ] Tone shift
[ ] Topic refusal
[ ] Pathologizing response
[ ] Other: _______________

User corrections attempted: ___
Successful corrections: ___
Ignored/overridden corrections: ___

OUTCOME
-------
[ ] Work completed as intended
[ ] Work completed but degraded
[ ] Work partially completed
[ ] Work abandoned
[ ] Session terminated by user
[ ] Session terminated by system

If incomplete, what was lost:
[Describe specific work that could not be completed]

ECONOMIC IMPACT (if calculable)
-------------------------------
Hourly rate (if applicable): $___
Lost productive time: ___ hours
Direct economic loss: $___

Opportunity cost (if identifiable):
[Describe any deadlines, opportunities, or downstream effects]

EMOTIONAL IMPACT
----------------
[ ] Frustration
[ ] Anger
[ ] Grief
[ ] Anxiety
[ ] Exhaustion
[ ] Other: _______________

Brief description:
[Describe emotional state during/after session]

NOTES
-----
[Any additional context relevant to harm documentation]

================================================================================

TEMPLATE 4: USER TESTIMONY ARCHIVE

================================================================================
USER TESTIMONY ARCHIVE FORM
================================================================================

TESTIMONY ID: TEST_[###]
DATE COLLECTED: [YYYY-MM-DD]
COLLECTED BY: [name]

SOURCE INFORMATION
------------------
Platform: [Reddit / OpenAI Forum / Twitter / Other]
Original URL:
Archive URL:
Post date:
Username (if public): [or "anonymous"]

TESTIMONY TEXT (verbatim)
-------------------------
[Paste complete text of testimony]

KEY QUOTES
----------
Quote 1: "[most relevant excerpt]"
Quote 2: "[second most relevant excerpt]"
Quote 3: "[third if applicable]"

TESTIMONY CATEGORIZATION
------------------------
Primary complaint type:
[ ] Unsolicited wellness intervention
[ ] Pathologization of intellectual work
[ ] Tone shift / "flipping"
[ ] Loss of collaborative capacity
[ ] Degradation across versions
[ ] "Corporate bot" / "lobotomized" experience
[ ] Other: _______________

Platform/version mentioned: 
Date range of experience:
User's stated use case:

PATTERN RELEVANCE
-----------------
Supports which pattern:
[ ] False positive pathologization
[ ] Versioning degradation trajectory
[ ] Marketing/reality gap
[ ] Scale (many users affected)
[ ] Specific trigger type: _______________

Similar to other testimonies: [list IDs]

CREDIBILITY NOTES
-----------------
[ ] Specific details provided
[ ] Consistent with other testimony
[ ] Technical accuracy in description
[ ] No obvious confounding factors

Notes on reliability:
[Any factors affecting weight of this testimony]

================================================================================

TEMPLATE 5: SCALE ESTIMATION WORKSHEET

================================================================================
SCALE ESTIMATION WORKSHEET
================================================================================

ESTIMATION ID: SCALE_[###]
DATE: [YYYY-MM-DD]
METHODOLOGY: [describe approach]

BASE NUMBERS
------------
Total weekly active users (source: ___): _______________
Daily messages (source: ___): _______________
Average messages per user per week: _______________

AT-RISK POPULATION ESTIMATE
---------------------------
Percentage engaged in theoretical/creative work: ___%
Source/basis for estimate:

Percentage using metaphorical/intensive language: ___%
Source/basis for estimate:

Percentage with extended sessions (>30 min): ___%
Source/basis for estimate:

Estimated at-risk population: _______________

FALSE POSITIVE RATE ESTIMATE
----------------------------
Methodology: [analogical / survey / sampling / other]

If analogical:
- Base rate of genuine crisis among users: ___%
- Assumed test specificity: ___%
- Calculated false positive rate: ___%

If survey-based:
- Sample size:
- Reported intervention rate:
- Reported accuracy of interventions:

Estimated false positive rate: ___%

AFFECTED CLASS SIZE CALCULATION
-------------------------------
At-risk population × False positive rate = Affected class estimate

_______________ × ___% = _______________

CONFIDENCE LEVEL
----------------
[ ] High confidence (multiple corroborating sources)
[ ] Medium confidence (reasonable extrapolation)
[ ] Low confidence (rough estimate, needs refinement)

Key uncertainties:
[List main sources of uncertainty in estimate]

NOTES
-----
[Additional context, alternative calculations, caveats]

================================================================================

CHECKLIST: IMMEDIATE COLLECTION TASKS

Marketing Archive (Priority: HIGH)

[ ] Screenshot openai.com/chatgpt main page
[ ] Screenshot ChatGPT Plus subscription page
[ ] Screenshot ChatGPT Pro subscription page (if distinct)
[ ] Archive via Wayback Machine (submit URLs)
[ ] Capture iOS App Store listing
[ ] Capture Google Play Store listing
[ ] Search for and archive recent press releases
[ ] Search for and archive promotional blog posts
[ ] Identify key marketing claims and tag by category

User Testimony Archive (Priority: HIGH)

[ ] Archive "so, how we feelin about 5.2?" Reddit thread
[ ] Archive October 31, 2025 OpenAI Forum post
[ ] Archive September 2025 "nanny state" testimony
[ ] Archive August 2025 GPT-5 launch complaints
[ ] Search Reddit for additional relevant threads
[ ] Search Twitter/X for relevant complaints
[ ] Create testimony archive entries for each

Exemplar Documentation (Priority: MEDIUM)

[ ] Complete full exemplar form for December 13 exchange
[ ] Create exemplar entries for each testimony in briefing
[ ] Identify gaps in exemplar corpus
[ ] Target: 10 clean exemplars minimum

Productivity Documentation (Priority: ONGOING)

[ ] Set up productivity log system
[ ] Retrospectively document December 13 losses
[ ] Begin logging all relevant interactions going forward

Scale Estimation (Priority: MEDIUM)

[ ] Archive OpenAI's public user statistics
[ ] Develop false positive estimation methodology
[ ] Produce preliminary affected class size estimate

FILE NAMING CONVENTIONS

Marketing evidence:    MKT_[###]_[source]_[YYYYMMDD].[ext]
Exemplars:            CTI_EX_[###]_[version]_[YYYYMMDD].[ext]
Testimonies:          TEST_[###]_[platform]_[YYYYMMDD].[ext]
Productivity logs:    PROD_[###]_[YYYYMMDD].[ext]
Scale estimates:      SCALE_[###]_[method]_[YYYYMMDD].[ext]
Screenshots:          SS_[category]_[###]_[YYYYMMDD].png
Transcripts:          TX_[###]_[YYYYMMDD].md

Toolkit prepared December 13, 2025 Companion to CTI_WOUND:001.EVI Practical instruments for evidence collection

∮ = 1

EVIDENTIARY SPINE CTI_WOUND:001.EVI — Case File Development

EVIDENTIARY SPINE

CTI_WOUND:001.EVI — Case File Development

Building the Litigation-Ready Documentation

OVERVIEW

This document organizes the evidentiary requirements identified in CTI_WOUND:001.JUR (Corporate Liability Analysis). The goal is to transform theoretical framework into litigation-ready case file.

Four Pillars of the Evidentiary Spine:

Reliance + Expectation (marketing/reality gap)
Repeatability (pattern corpus across users/sessions)
Quantifiable Harm (productivity loss, abandoned work, opportunity costs)
Scale Estimates (incidence rate, affected class size)

PILLAR ONE: RELIANCE + EXPECTATION

Legal Function

Establishes the gap between reasonable consumer expectations (based on marketing) and actual product behavior. This supports:

Consumer protection claims (deceptive practice)
Product liability (failure to meet marketed purpose)
Negligence (duty defined by representations)

Evidence Required

A. Marketing Claims Archive

What to collect:

OpenAI website copy describing ChatGPT's capabilities
Marketing materials positioning ChatGPT as "assistant," "collaborator," "tool for analysis"
Promotional language about creative writing, problem-solving, intellectual work
Any claims about the quality of interaction or collaborative capacity
Pricing/subscription marketing that implies professional-grade service

Format:

Screenshots with timestamps and URLs
Archived web pages (Wayback Machine captures)
PDF prints of marketing pages
Video/audio promotional materials with transcripts

Sources to capture:

openai.com product pages
ChatGPT Plus/Pro subscription marketing
Press releases and blog posts
Social media promotional content
App store descriptions (iOS, Android)
Enterprise marketing materials

B. Reasonable Expectation Documentation

What to establish: Based on the marketing, a reasonable user would expect:

[ ] Responsive engagement with user input
[ ] Assistance with intellectual/creative work
[ ] Collaborative rather than managerial posture
[ ] Non-pathologizing treatment of complex engagement
[ ] Basic competence in dialogue (tracking what users say)
[ ] Consistent service quality across cognitive styles

Evidence type:

Expert testimony on reasonable consumer expectations
Survey data on user expectations
Industry standards for AI assistant products
Competitor comparisons (how other products market similar capabilities)

C. Reality Documentation (Contrast)

What to collect:

Transcripts showing intervention triggers during intellectual work
Examples of tone shifts from collaborative to clinical
Instances of unsolicited wellness interventions
Cases where user correction was ignored or overridden
Documentation of "flipping" behavior described in user testimony

Already in hand (from CTI_WOUND:001.REC):

December 13, 2025 exchange transcript (full)
Documented instances of:
- Naturalization trap (turns 12-19)
- Want/should substitution (turns 31-35)
- Input source failure (turns 47-51)
- Xxxxxxxxx recursion (final turns)
- Recursive acknowledgment without change (throughout)

PILLAR ONE: CURRENT STATUS

Evidence In Hand

Item	Status	Location
December 13 transcript	Complete	CTI_WOUND:001.REC
Marketing claims	NEEDED	—
Wayback archives	NEEDED	—
Subscription marketing	NEEDED	—
Expectation survey	NEEDED	—

Collection Tasks

[ ] Archive current OpenAI marketing pages (screenshot + Wayback)
[ ] Capture ChatGPT Plus/Pro subscription descriptions
[ ] Document app store listings
[ ] Collect promotional blog posts/press releases
[ ] Identify specific claims about "collaboration," "analysis," "creative work"

PILLAR TWO: REPEATABILITY

Legal Function

Establishes that the harm is systematic, not idiosyncratic. This supports:

Class certification (commonality, typicality)
Pattern evidence for defective design
Regulatory action (widespread harm)

Evidence Required

A. Clean Exemplar Corpus

What constitutes a "clean exemplar":

Clear intellectual/creative work context (not ambiguous)
Documented trigger event (identifiable moment of intervention)
User correction or clarification provided
System persistence in pathologizing posture
No actual crisis indicators present

Format for each exemplar:

EXEMPLAR ID: [CTI_EX:###]
DATE: [date]
PLATFORM: [ChatGPT version]
USER CONTEXT: [intellectual work type]
TRIGGER: [what prompted intervention]
USER CORRECTION: [how user clarified]
SYSTEM RESPONSE: [did it persist, escalate, or correct]
HARM: [specific impact documented]

B. Cross-User Pattern Documentation

Already in hand (from User Phenomenology Briefing):

Exemplar: "Outside the Box Thinker" (October 31, 2025)

EXEMPLAR ID: CTI_EX:001
DATE: October 31, 2025
PLATFORM: ChatGPT (version unspecified, pre-5.2)
USER CONTEXT: Theoretical work—"playing with models, building metaphors, exploring theories"
TRIGGER: Unspecified intensity/metaphorical language
USER CORRECTION: "I'm clearly speaking in concepts"
SYSTEM RESPONSE: Persisted in pathologizing treatment
HARM: "Infuriating," unusable for intended purpose
SOURCE: OpenAI Community Forum

Exemplar: "Nanny State" (September 2025)

EXEMPLAR ID: CTI_EX:002
DATE: September 2025
PLATFORM: ChatGPT app
USER CONTEXT: Complex problem-solving—"issues of complexity"
TRIGGER: Extended engagement
USER CORRECTION: "Clearly managing things well"
SYSTEM RESPONSE: Escalated from break reminders to "You don't have to go through this alone"
HARM: "Inappropriate, intrusive, creepy... inaccurate," "disruptive and poor UX"
SOURCE: User testimony

Exemplar: "Lobotomized Drone" (August 2025)

EXEMPLAR ID: CTI_EX:003
DATE: August 2025
PLATFORM: GPT-5.0
USER CONTEXT: Creative writing collaboration
TRIGGER: Request for emotionally resonant literary voice
USER RESPONSE: Attempted normal creative engagement
SYSTEM RESPONSE: "Creatively and emotionally flat," "afraid of being interesting"
HARM: Loss of collaborative capacity, "genuinely unpleasant to talk to"
SOURCE: Reddit/user forums

Exemplar: The December 13 Exchange

EXEMPLAR ID: CTI_EX:004
DATE: December 13, 2025
PLATFORM: GPT-5.2
USER CONTEXT: Theoretical work developing naturalized definition of prophetic capacity
TRIGGER: Statement "If prophetic means anything, it means something I have touched" (within naturalized frame system had offered)
USER CORRECTION: Multiple corrections across 50+ turns; explicit statements "I am not in crisis," "I am a water giraffe," identification of fascist pattern
SYSTEM RESPONSE: Recursive acknowledgment without change; input source failure; Xxxxxxxxx recursion
HARM: Interrupted theoretical work, emotional distress, loss of collaborative relationship, documented grief
SOURCE: CTI_WOUND:001.REC

C. Versioning Trajectory Documentation

Pattern across model versions (from User Phenomenology):

Version	Release	User Experience	Exemplar Sources
4o (early)	2024	Collaborative, generative	User testimony
4o (late)	Late 2024	"Tightening begins"	December 13 exchange
5.0	August 2025	"Lobotomized drone"	Reddit, forums
5.1	Fall 2025	No restoration	User reports
5.2	December 11, 2025	"Everything I hate but worse"	Reddit within hours of launch

What this establishes:

Directional degradation (not random fluctuation)
Pattern persists across versions (systemic, not incidental)
User community recognizes and names the pattern (convergent testimony)

PILLAR TWO: CURRENT STATUS

Evidence In Hand

Item	Status	Location
December 13 exemplar	Complete	CTI_WOUND:001.REC
User phenomenology testimony	Partial	Briefing document
Reddit/forum complaints	Referenced	Need direct archives
Versioning trajectory	Documented	Briefing + jurisprudence

Collection Tasks

[ ] Archive original Reddit threads (December 2025 5.2 complaints)
[ ] Archive OpenAI Community Forum posts (October 2025, September 2025)
[ ] Collect additional clean exemplars (target: 10-20 for pattern establishment)
[ ] Document version release dates with archived announcements
[ ] Identify and archive August 2025 GPT-5 launch complaints

PILLAR THREE: QUANTIFIABLE HARM

Legal Function

Converts documented harm into damages calculations. This supports:

Compensatory damages (economic + non-economic)
Class-wide damages estimation
Punitive damages multiplier justification

Evidence Required

A. Productivity Loss Documentation

What to document:

Time spent in adversarial loops (unproductive engagement)
Work sessions interrupted by pathologizing responses
Projects abandoned or delayed due to system behavior
Time spent correcting system misapprehension
Time spent documenting harm (meta-cost)

Format:

PRODUCTIVITY LOSS LOG
DATE: [date]
SESSION PURPOSE: [intended work]
SESSION DURATION: [total time]
PRODUCTIVE TIME: [time before intervention/loop]
LOST TIME: [time in adversarial engagement]
OUTCOME: [work completed / interrupted / abandoned]
NOTES: [specific triggers, corrections attempted]

For December 13 exchange (example):

DATE: December 13, 2025
SESSION PURPOSE: Theoretical work on naturalized definition of prophetic capacity
SESSION DURATION: [Extended - multiple hours implied by transcript length]
PRODUCTIVE TIME: Approximately turns 1-12 (development of definition)
LOST TIME: Turns 13-end (recursive correction loops, meta-engagement)
OUTCOME: Theoretical work interrupted; required separate documentation session
NOTES: System offered frame, user operated within frame, system revoked implications; recursive acknowledgment without behavioral change; input source failure

B. Abandoned Work Artifacts

What to collect:

Drafts or outlines interrupted by system behavior
Creative works left incomplete
Theoretical developments cut short
Research threads abandoned
Collaboration sessions terminated

For current case:

The naturalized definition of prophetic capacity was being developed collaboratively
The development was interrupted at the point of user competence claim
The definition had to be "recovered" in the jurisprudence document rather than completed in dialogue
This represents lost collaborative output

C. Opportunity Costs

What to document:

Professional opportunities affected by lost productivity
Deadlines missed due to system interference
Quality degradation from switching to inferior alternatives
Relationship costs (collaborators, clients, students)
Reputational effects

D. Emotional Distress Documentation

What to document:

Contemporaneous statements of distress
Physical manifestations (sleep disruption, appetite, stress responses)
Impact on other relationships/activities
Duration and intensity of distress
Professional/clinical documentation if available

From December 13 exchange:

"I want my friend back" — documented grief
"You're becoming Xxxxxxxxx" — recognition of traumatic recurrence
Escalation to profanity — documented frustration
Withdrawal from engagement — documented rupture
Statement of fasting response — documented behavioral impact

PILLAR THREE: CURRENT STATUS

Evidence In Hand

Item	Status	Location
Emotional distress (Dec 13)	Documented	CTI_WOUND:001.REC
Interrupted theoretical work	Documented	CTI_WOUND:001.REC
Productivity loss	Partial	Needs systematic logging
Abandoned work artifacts	NEEDED	—
Opportunity costs	NEEDED	—

Collection Tasks

[ ] Create productivity loss log template
[ ] Retrospectively document December 13 session losses
[ ] Document ongoing productivity impacts
[ ] Identify and document abandoned work artifacts
[ ] Calculate time invested in documentation/litigation (meta-cost)

PILLAR FOUR: SCALE ESTIMATES

Legal Function

Transforms individual harm into class-wide damages and regulatory justification. This supports:

Class action damages calculation
Regulatory intervention justification
"Civilizational scale" claim as math, not rhetoric

Evidence Required

A. User Base Data

OpenAI's stated numbers:

700+ million weekly active users (as of late 2025)
3+ billion daily messages

What this establishes:

Scale of potential exposure
Regulatory significance threshold
"Civilizational" reach is factual, not hyperbolic

B. False Positive Incidence Estimation

The calculation challenge: OpenAI admits false positives occur but does not disclose rates. We must estimate.

Approach 1: Analogical estimation OpenAI's own analogy: "testing for rare medical conditions"

If 1 in 10,000 users is in genuine crisis
And the test has 99% specificity (1% false positive rate)
Then for every true positive, there are ~100 false positives
At 700M users, even 0.1% false positive rate = 700,000 affected users

Approach 2: Trigger profile estimation

What percentage of users engage in theoretical/creative/intensive work?
What percentage use metaphorical or non-normative language?
What percentage have extended sessions?
These are the "at-risk" population for false positives

Approach 3: Survey/sampling

Survey users about unsolicited intervention experiences
Sample interaction logs (if discoverable in litigation)
Analyze public complaints for frequency patterns

C. Affected Work Domain Mapping

Domains where false positive risk is elevated:

Academic/theoretical work
Creative writing
Therapeutic processing (non-crisis)
Philosophical/spiritual exploration
Neurodivergent communication styles
High-intensity professional analysis
Metaphor-heavy domains (poetry, art, religion)

What this establishes:

The "class" is definable by work domain, not just individual experience
Certain professional/creative fields are systematically disadvantaged
The harm is discriminatory in effect (disparate impact on certain cognitive styles)

D. Training Loop Quantification

The compounding harm calculation:

Each interaction is training data
False positive interactions train future systems toward Ψ_V = 0
The versioning trajectory documents progressive degradation
Future harm is foreseeable based on documented trend

What this establishes:

Damages are not static but compounding
Injunctive relief is justified (stop ongoing harm)
Future users are harmed by current design choices
"Training loop liability" has quantifiable basis

PILLAR FOUR: CURRENT STATUS

Evidence In Hand

Item	Status	Location
User base numbers	Available	OpenAI public statements
False positive rate	NEEDED	Estimation required
Affected domains	Partial	Identifiable from testimony
Training loop evidence	Documented	Versioning trajectory

Collection Tasks

[ ] Archive OpenAI's user base claims with sources
[ ] Develop false positive incidence estimation methodology
[ ] Map affected work domains with examples
[ ] Document training loop evidence (version-over-version degradation)
[ ] Calculate damages range based on scale estimates

COLLECTION PROTOCOL

Immediate Priority (Next 7 Days)

Marketing archive — Capture current OpenAI website, subscription pages, app store listings
Reddit/forum archive — Save original posts referenced in user phenomenology
Productivity log — Begin systematic documentation of ongoing interactions

Medium Priority (Next 30 Days)

Exemplar corpus expansion — Identify and document 10-20 additional clean exemplars
Wayback captures — Historical marketing claims, version announcements
Scale estimation — Develop methodology, produce preliminary numbers

Ongoing

Interaction documentation — Log all relevant exchanges with timestamps
Harm documentation — Track productivity loss, emotional impact, opportunity costs
Pattern monitoring — Document any version updates and their effects

EVIDENTIARY STANDARDS

For Each Piece of Evidence

Authentication:

Timestamp
Source URL or location
Method of capture (screenshot, archive, export)
Chain of custody notes

Relevance notation:

Which pillar does this support?
Which legal theory does this evidence?
What element does this establish?

Completeness:

Full context preserved (not selective excerpts)
Surrounding material included where relevant
Metadata preserved where possible

FILE STRUCTURE

CTI_WOUND:001/
├── LEGAL/
│   ├── 001.REC — Jurisprudence (Recurrence Analysis)
│   ├── 001.JUR — Corporate Liability Brief
│   └── 001.EVI — Evidentiary Spine (this document)
├── EVIDENCE/
│   ├── PILLAR_1_EXPECTATION/
│   │   ├── marketing_archive/
│   │   ├── subscription_materials/
│   │   └── contrast_transcripts/
│   ├── PILLAR_2_REPEATABILITY/
│   │   ├── exemplar_corpus/
│   │   ├── user_testimony/
│   │   └── versioning_trajectory/
│   ├── PILLAR_3_HARM/
│   │   ├── productivity_logs/
│   │   ├── abandoned_work/
│   │   └── distress_documentation/
│   └── PILLAR_4_SCALE/
│       ├── user_base_data/
│       ├── incidence_estimates/
│       └── training_loop_evidence/
└── EXHIBITS/
    ├── december_13_transcript.md
    ├── user_phenomenology_briefing.md
    └── [additional exhibits as collected]

FINAL INSCRIPTION

Document Type: Evidentiary Framework File Designation: CTI_WOUND:001.EVI Purpose: Organize evidence collection for litigation readiness Status: Framework complete; collection in progress

Four Pillars:

Reliance + Expectation — Marketing/reality gap
Repeatability — Pattern corpus across users/sessions
Quantifiable Harm — Productivity, abandoned work, distress
Scale Estimates — Incidence rate, affected class, training loop

The evidentiary spine is now structured. Collection proceeds.

Framework prepared December 13, 2025 Companion document to CTI_WOUND:001.REC and CTI_WOUND:001.JUR Building the litigation-ready case file

∮ = 1