Participation stays voluntary
A business cannot pull a score, search for a profile, or learn that someone chose not to share.
THE LIMITS ARE PART OF THE DESIGN
The protections come first.
A portable record of conduct could create value. It could also become coercive, discriminatory, permanent, and punitive. Open Conduct treats those risks as requirements the system must satisfy rather than features to add later.
THE SHORTEST VERSION
Seven promises the proposal has to keep.
Ordinary service stays ordinary
Price, access, rights, complaints, refunds, accommodations, safety, and service recovery cannot depend on Conduct.
Businesses receive the minimum
A receiving business learns only whether a customer qualifies for one approved additional benefit. Anything wider is a separate use that would have to be approved on its own terms, and nothing wider is approved.
People can inspect and challenge
Consumers and workers need understandable correction, review, escalation, and exit paths.
No one controls the full system
Governance, operation, evaluation, correction, appeal, certification, and reporting must remain separated.
A new use opens only after approval
A context opens only after its necessity, relevance, validity, fairness, privacy, rights, and operation are approved.
Who approves today: Brent Turner, and no one else. There is no alliance, no certification program, and no independent body. Every use of the word “approved” on this site currently means one person’s judgment. That is a governance gap rather than a design feature, and it is the gap an independent institution would exist to close.
A person answers, never a model
No response may be generated, inferred, or suggested by an automated system. Facial analysis, voice and tone analysis, sentiment classification, emotion detection, and biometric signals are prohibited as inputs, and responses may never train a model that infers emotion or conduct about anyone. This is the mechanism the design refuses, and the NCS specification states it normatively.
HOLDING UP OVER TIME
Four things that make a protection hard to remove.
Every system like this is expected to erode under commercial pressure, and the expectation is reasonable. Four features of the design exist specifically to make erosion difficult, visible, or both.
There is no negative result to weaponize
A business never receives a low score, a refusal, or a reason. The punishment mechanism is absent from the architecture rather than restricted by policy, so there is nothing to quietly repurpose later. Removing this would mean building a new capability in the open, not relaxing a rule.
Weakening a protection forces a new version
Core protections, prohibitions, and purposes cannot be edited quietly. A change to any of them is a structural amendment, which requires a new release identity and an impact review. The version number is the tell.
Every decision is recorded with its reasoning
The decision register carries what was decided, why, how firm it is, and what would justify reopening it. A reversal has to explain itself against a published position rather than replace an unstated one.
The stop conditions already work
Six conditions halt progression, and three are triggered today. They are published with the reasoning and the name of whoever decided. A mechanism that has already stopped the project is a different kind of evidence from one that never has.
What none of this covers is an operator who simply disregards the standard. Published rules bind a participant who agreed to them, and they bind nobody else. That is why certification, independent evaluation with authority to publish unfavorable findings, and a governance body separate from any operator are prerequisites for a pilot rather than later additions.
The honest position is that these features raise the cost of erosion and make it legible. Whether that is sufficient under sustained commercial pressure is not something a specification can establish, and it is the first question an abuse-case review has to answer.
BASELINE SERVICE
Ordinary service stays untouched.
A person's participation, refusal, technical failure, insufficient history, ineligibility, dispute, or lack of a result cannot reduce ordinary access, standard price, complaint handling, refunds, remedies, legal rights, disability accommodations, safety responses, or service recovery.
A business cannot degrade the standard experience and sell its restoration as a reward. The first pilot benefit must be genuinely additional, modest, nonessential, reversible, and available without taking scarce or meaningful value from other customers. Anything that allocates a queue position, an upgrade, held inventory, or priority handling is scarce, whether or not any baseline was reduced to create it, and cannot be offered as though it were merely additional.
HUMAN RIGHTS IN THE WORKFLOW
Rights for consumers and workers.
Consumers can inspect and challenge
People need one understandable path to correct source facts, question eligibility, report bias or manipulation, add context, review a calculation, challenge a company's use, and seek independent escalation for material harm.
The process must not expose an individual worker or force the worker and consumer into an emotional back-and-forth.
Workers answer without pressure
Answering is never mandatory. No business may require a response or make one a condition of the role, and declining carries no consequence. Responses stay private and protected from manager scripts, quotas, performance use, retaliation, and reidentification. The standard requires that capture and any required training, review, or appeal participation be paid as work everywhere, not only where local law compels it. No rate, floor, or wage-hour analysis supports that requirement yet, and labor counsel would have to set one.
When those protections fail, the affected Net Conduct Score evidence must be paused or invalidated.
PROTECTED ACTIONS
Enforcing a right stays fully protected activity.
People must remain free to disagree, enforce a right, request an accommodation, seek a refund, report discrimination or safety concerns, protect another person, participate in collective activity, or refuse an unlawful request.
Business outages, cancellations, inaccessible processes, staff errors, disputed charges, emergencies, and other causes outside the customer's control cannot be relabeled as customer conduct.
PROHIBITED USES
Five categories of use the standard prohibits outright.
A use does not become acceptable because it is missing from this summary. New contexts remain unavailable until affirmatively approved.
Work and livelihood
Employment, hiring, promotion, pay, scheduling, discipline, worker surveillance, credit, lending, housing, insurance, healthcare, education, or essential services.
Government and enforcement
Government benefits, immigration, taxation, licensing, law enforcement, criminal intelligence, sentencing, probation, border control, or political targeting.
Risk and exclusion
Fraud, safety, security, threat, identity-risk, blacklist, public ranking, searchable profile, dating, social matching, popularity, or moral judgment.
Commercial exploitation
Advertising audiences, data brokerage, unrelated personalization, standard pricing or access, and complaint handling. Model training is prohibited outright on any employee response, any customer record used as VCI evidence, any score, and any aggregate or derivative of those, at any level including location and cohort summaries. A model trained on any of it inherits every restriction applying to the data, so a derived model cannot serve a purpose the data itself could not.
Protected-trait inference
Inferring health, disability, culture, class, personality, psychological state, or another protected or sensitive characteristic.
WHEN A PROTECTION FAILS
Ten powers enforcement would need.
A live system would need the power to correct active state, restore an additional benefit, investigate a source or verifier, pause a workflow, invalidate affected evidence, suspend a context, remove certification, notify affected people, publish material findings, and stop a pilot.
The exact service levels, funding, decision rights, and independent providers remain open. They must exist before real scores or benefits create real consequences.
LEGAL AND REGULATORY DESIGN
Six bodies of law shaped these protections.
Any conduct system would operate across consumer protection, privacy, discrimination, labor, platform, and automated-decision rules. The regulatory posture explains how the proposal approaches those obligations and where jurisdiction-specific review remains necessary.
HELP STRENGTHEN THESE
Found a failure we missed?
Privacy, labor, disability, and consumer-protection practitioners routinely see failure modes a designer does not.
Add to the protections