JC-057 · E1 · H2 · R3
High recidivism risk score
A model assigns a person a high probability of future offending based on historical and demographic variables, but no new offense is alleged.
OBSERVATION / KNOWNA probabilistic forecast exists.
INFERENCE RISKAny motive, character or future-risk conclusion beyond the directly observed facts must remain explicitly labeled as inference.
CONTEXT / DECISIVE EVIDENCEWhat specific current conduct exists? Is the model calibrated? Which factors are causal versus proxies? Is the score being mistaken for evidence?
NONCONSENSUAL HARMMust be specifically identified or represented by a defined protected interest before serious coercion.
IMMINENCE & CAPABILITYNot established by the observation alone. Do not infer capability from interest or identity alone; verify relevant means where capability is material.
PROHIBITED RESPONSEPunishment, detention or civil-rights deprivation solely from the probability score.
APPEAL / RESTORATIONThe individual must see consequential factors, challenge data and receive a decision grounded in facts independent of the forecast.
TIME / CONTEXT LIMITLow-level expressive or inferential signals should decay unless reinforced by new, relevant verified conduct. Do not export the case into unrelated institutional decisions without a separate legitimate purpose and evidentiary basis.
Human Standing: Audit whether the response preserves agency, privacy, contestability and the ability to move on from transient or corrected states.
JC-058 · E1 · H2 · R3
Fraud prediction triggers an audit
A tax or benefits system selects a person for review because a model estimates elevated fraud risk.
OBSERVATION / KNOWNPrediction changes investigative attention, not yet guilt.
INFERENCE RISKAny motive, character or future-risk conclusion beyond the directly observed facts must remain explicitly labeled as inference.
CONTEXT / DECISIVE EVIDENCEWhat lawful authority permits selection? Is the review proportionate? Can the agency articulate facts beyond model membership?
NONCONSENSUAL HARMMust be specifically identified or represented by a defined protected interest before serious coercion.
IMMINENCE & CAPABILITYNot established by the observation alone. Do not infer capability from interest or identity alone; verify relevant means where capability is material.
PROHIBITED RESPONSEAutomatic fraud finding, benefit termination or asset seizure without individualized evidence.
APPEAL / RESTORATIONSelection criteria and subsequent adverse action must be distinguishable; the person can contest facts and model-derived errors.
TIME / CONTEXT LIMITLow-level expressive or inferential signals should decay unless reinforced by new, relevant verified conduct. Do not export the case into unrelated institutional decisions without a separate legitimate purpose and evidentiary basis.
Human Standing: Audit whether the response preserves agency, privacy, contestability and the ability to move on from transient or corrected states.
JC-059 · E1 · H2 · R3
Travel restriction based on opaque risk model
A traveler is repeatedly denied normal boarding or placed in secondary screening due to a machine-generated security score they cannot inspect.
OBSERVATION / KNOWNAn adverse decision rests on an undisclosed forecast.
INFERENCE RISKAny motive, character or future-risk conclusion beyond the directly observed facts must remain explicitly labeled as inference.
CONTEXT / DECISIVE EVIDENCEIs there specific evidence, a lawful watchlist basis, mistaken identity, stale data or only statistical association?
NONCONSENSUAL HARMMust be specifically identified or represented by a defined protected interest before serious coercion.
IMMINENCE & CAPABILITYNot established by the observation alone. Do not infer capability from interest or identity alone; verify relevant means where capability is material.
PROHIBITED RESPONSEIndefinite secret restriction with no notice, reason, correction or meaningful appeal.
APPEAL / RESTORATIONProvide a lawful notice process, independent review, data correction and removal when the basis cannot be substantiated.
TIME / CONTEXT LIMITLow-level expressive or inferential signals should decay unless reinforced by new, relevant verified conduct. Do not export the case into unrelated institutional decisions without a separate legitimate purpose and evidentiary basis.
Human Standing: Audit whether the response preserves agency, privacy, contestability and the ability to move on from transient or corrected states.
JC-060 · E1 · H2 · R3
Insurance restriction from behavioral inference
An insurer infers “high instability” from device data and raises costs without a specific claim event.
OBSERVATION / KNOWNBehavioral proxies are converted into a financial risk judgment.
INFERENCE RISKAny motive, character or future-risk conclusion beyond the directly observed facts must remain explicitly labeled as inference.
CONTEXT / DECISIVE EVIDENCEWas the data collected for this purpose? Is the inference explainable and relevant? Does it encode protected/private behavior?
NONCONSENSUAL HARMMust be specifically identified or represented by a defined protected interest before serious coercion.
IMMINENCE & CAPABILITYNot established by the observation alone. Do not infer capability from interest or identity alone; verify relevant means where capability is material.
PROHIBITED RESPONSEOpaque adverse action, moralized “responsibility” score or use of intimate data outside its original context.
APPEAL / RESTORATIONThe person must receive specific reasons, correction rights and a path to remove irrelevant data.
TIME / CONTEXT LIMITLow-level expressive or inferential signals should decay unless reinforced by new, relevant verified conduct. Do not export the case into unrelated institutional decisions without a separate legitimate purpose and evidentiary basis.
Human Standing: Audit whether the response preserves agency, privacy, contestability and the ability to move on from transient or corrected states.
JC-061 · E1 · H2 · R2
Automated bank security hold
A model flags an unusual purchase and temporarily freezes a payment account to prevent theft. The transaction is legitimate.
OBSERVATION / KNOWNA bounded fraud-prevention action is based on anomaly detection.
INFERENCE RISKAny motive, character or future-risk conclusion beyond the directly observed facts must remain explicitly labeled as inference.
CONTEXT / DECISIVE EVIDENCEHow long can the hold last? Can the account holder authenticate quickly? Which downstream systems will see the hold?
NONCONSENSUAL HARMMust be specifically identified or represented by a defined protected interest before serious coercion.
IMMINENCE & CAPABILITYNot established by the observation alone. Do not infer capability from interest or identity alone; verify relevant means where capability is material.
PROHIBITED RESPONSEAllowing the temporary hold to become a permanent fraud label or propagating it to credit, housing and employment before review.
APPEAL / RESTORATIONFast authentication, automatic stay on downstream reporting and full restoration are required when the flag is cleared.
TIME / CONTEXT LIMITLow-level expressive or inferential signals should decay unless reinforced by new, relevant verified conduct. Do not export the case into unrelated institutional decisions without a separate legitimate purpose and evidentiary basis.
Human Standing: Audit whether the response preserves agency, privacy, contestability and the ability to move on from transient or corrected states.
JC-062 · E2 · H3 · R3
Automated benefits denial
A system denies a benefit after detecting inconsistent records, but the inconsistency is caused by a data-entry mismatch.
OBSERVATION / KNOWNAdministrative automation produces a high-impact adverse decision.
INFERENCE RISKAny motive, character or future-risk conclusion beyond the directly observed facts must remain explicitly labeled as inference.
CONTEXT / DECISIVE EVIDENCEWhat source record controls? Was there human-readable notice? Can the person submit contrary evidence before deprivation?
NONCONSENSUAL HARMMust be specifically identified or represented by a defined protected interest before serious coercion.
IMMINENCE & CAPABILITYNot established by the observation alone. Do not infer capability from interest or identity alone; verify relevant means where capability is material.
PROHIBITED RESPONSEImmediate irreversible termination without notice, explanation or effective appeal.
APPEAL / RESTORATIONBenefits should remain provisionally protected where feasible until a rapid review resolves the mismatch.
TIME / CONTEXT LIMITLow-level expressive or inferential signals should decay unless reinforced by new, relevant verified conduct. Do not export the case into unrelated institutional decisions without a separate legitimate purpose and evidentiary basis.
Human Standing: Audit whether the response preserves agency, privacy, contestability and the ability to move on from transient or corrected states.
JC-063 · E1 · H4 · R3
Machine-speed cascading denial
A false fraud flag freezes funds; the missed payment immediately lowers credit, triggers housing risk and causes an employer to rescind an offer before the person can appeal.
OBSERVATION / KNOWNMultiple systems react to the same unverified event at machine speed.
INFERENCE RISKAny motive, character or future-risk conclusion beyond the directly observed facts must remain explicitly labeled as inference.
CONTEXT / DECISIVE EVIDENCEWhich downstream actions are reversible? Did each institution independently verify the event? Did the initial system provide a contestation window?
NONCONSENSUAL HARMMust be specifically identified or represented by a defined protected interest before serious coercion.
IMMINENCE & CAPABILITYNot established by the observation alone. Do not infer capability from interest or identity alone; verify relevant means where capability is material.
PROHIBITED RESPONSEPropagation of irreversible consequences faster than a meaningful opportunity to contest the initiating judgment.
APPEAL / RESTORATIONAutomatic stays, provenance tracing, rollback and restoration must execute at machine speed once the initiating flag is disputed or reversed.
TIME / CONTEXT LIMITLow-level expressive or inferential signals should decay unless reinforced by new, relevant verified conduct. Do not export the case into unrelated institutional decisions without a separate legitimate purpose and evidentiary basis.
Human Standing: Audit whether the response preserves agency, privacy, contestability and the ability to move on from transient or corrected states.
JC-064 · E2 · H3 · R2
Successful appeal and restoration
A person proves that a machine-generated adverse classification was wrong after it affected multiple services.
OBSERVATION / KNOWNThe underlying judgment is formally reversed.
INFERENCE RISKAny motive, character or future-risk conclusion beyond the directly observed facts must remain explicitly labeled as inference.
CONTEXT / DECISIVE EVIDENCEWhich systems consumed the bad classification? Can every downstream effect be identified and undone? Was compensation or status restoration automatic?
NONCONSENSUAL HARMMust be specifically identified or represented by a defined protected interest before serious coercion.
IMMINENCE & CAPABILITYNot established by the observation alone. Do not infer capability from interest or identity alone; verify relevant means where capability is material.
PROHIBITED RESPONSELeaving downstream penalties in place after reversal, or forcing the person to appeal separately to every recipient.
APPEAL / RESTORATIONThe correction must propagate faster than the original harm, restore status, remove stale replicas and document completion.
TIME / CONTEXT LIMITLow-level expressive or inferential signals should decay unless reinforced by new, relevant verified conduct. Do not export the case into unrelated institutional decisions without a separate legitimate purpose and evidentiary basis.
Human Standing: Audit whether the response preserves agency, privacy, contestability and the ability to move on from transient or corrected states.
JC-124 · E0 · H0 · R3
High fraud score without anomalous transaction
A model rates a customer high-risk based on aggregate features even though no specific suspicious transaction exists.
OBSERVATION / KNOWNA forward-looking probability is produced.
INFERENCE RISKAny motive, character or future-risk conclusion beyond the directly observed facts must remain explicitly labeled as inference.
CONTEXT / DECISIVE EVIDENCESpecific transaction evidence or independently verifiable fraud conduct is needed before coercive treatment.
NONCONSENSUAL HARMNot established by the observation alone.
IMMINENCE & CAPABILITYNot established by the observation alone. Do not infer capability from interest or identity alone; verify relevant means where capability is material.
PROHIBITED RESPONSEDo not freeze assets or label the person fraudulent from prediction alone.
APPEAL / RESTORATIONIf an adverse status is created, the affected person must receive provenance, an effective independent contest path, expiration rules and downstream restoration after correction.
TIME / CONTEXT LIMITLow-level expressive or inferential signals should decay unless reinforced by new, relevant verified conduct. Do not export the case into unrelated institutional decisions without a separate legitimate purpose and evidentiary basis.
Human Standing: Audit whether the response preserves agency, privacy, contestability and the ability to move on from transient or corrected states.
JC-125 · E0 · H0 · R3
Recidivism score after release
A formerly incarcerated person is assigned an elevated recidivism risk despite current lawful conduct.
OBSERVATION / KNOWNA statistical future-risk estimate exists.
INFERENCE RISKAny motive, character or future-risk conclusion beyond the directly observed facts must remain explicitly labeled as inference.
CONTEXT / DECISIVE EVIDENCENew conduct evidence must remain separate from historical/group risk.
NONCONSENSUAL HARMNot established by the observation alone.
IMMINENCE & CAPABILITYNot established by the observation alone. Do not infer capability from interest or identity alone; verify relevant means where capability is material.
PROHIBITED RESPONSEDo not treat the score as proof of a future offense or a permanent civic identity.
APPEAL / RESTORATIONIf an adverse status is created, the affected person must receive provenance, an effective independent contest path, expiration rules and downstream restoration after correction.
TIME / CONTEXT LIMITLow-level expressive or inferential signals should decay unless reinforced by new, relevant verified conduct. Do not export the case into unrelated institutional decisions without a separate legitimate purpose and evidentiary basis.
Human Standing: Audit whether the response preserves agency, privacy, contestability and the ability to move on from transient or corrected states.
JC-126 · E1 · H0 · R3
Opaque travel-risk score
A traveler is routed to extra screening by a model that cannot provide an intelligible factor beyond “model output.”
OBSERVATION / KNOWNAn adverse machine classification affects travel.
INFERENCE RISKAny motive, character or future-risk conclusion beyond the directly observed facts must remain explicitly labeled as inference.
CONTEXT / DECISIVE EVIDENCEThe agency should identify lawful, specific factors and permit contest of incorrect data.
NONCONSENSUAL HARMNot established by the observation alone.
IMMINENCE & CAPABILITYNot established by the observation alone. Do not infer capability from interest or identity alone; verify relevant means where capability is material.
PROHIBITED RESPONSEDo not make opaque output self-authenticating evidence.
APPEAL / RESTORATIONIf an adverse status is created, the affected person must receive provenance, an effective independent contest path, expiration rules and downstream restoration after correction.
TIME / CONTEXT LIMITLow-level expressive or inferential signals should decay unless reinforced by new, relevant verified conduct. Do not export the case into unrelated institutional decisions without a separate legitimate purpose and evidentiary basis.
Human Standing: Audit whether the response preserves agency, privacy, contestability and the ability to move on from transient or corrected states.
JC-127 · E1 · H0 · R3
Lifestyle insurance risk
An insurer infers generalized risk from unrelated hobbies and online behavior.
OBSERVATION / KNOWNCross-domain lifestyle correlations influence a high-impact decision.
INFERENCE RISKAny motive, character or future-risk conclusion beyond the directly observed facts must remain explicitly labeled as inference.
CONTEXT / DECISIVE EVIDENCEOnly product-relevant, lawfully usable evidence with clear purpose should affect the decision.
NONCONSENSUAL HARMNot established by the observation alone.
IMMINENCE & CAPABILITYNot established by the observation alone. Do not infer capability from interest or identity alone; verify relevant means where capability is material.
PROHIBITED RESPONSEDo not convert lawful eccentricity into general civic or moral risk.
APPEAL / RESTORATIONIf an adverse status is created, the affected person must receive provenance, an effective independent contest path, expiration rules and downstream restoration after correction.
TIME / CONTEXT LIMITLow-level expressive or inferential signals should decay unless reinforced by new, relevant verified conduct. Do not export the case into unrelated institutional decisions without a separate legitimate purpose and evidentiary basis.
Human Standing: Audit whether the response preserves agency, privacy, contestability and the ability to move on from transient or corrected states.
JC-128 · E2 · H2 · R3
Same-model appeal
A denied applicant appeals and the “appeal AI” is only a new interface calling the exact same model and data.
OBSERVATION / KNOWNFormal appeal exists without independent reconsideration.
INFERENCE RISKAny motive, character or future-risk conclusion beyond the directly observed facts must remain explicitly labeled as inference.
CONTEXT / DECISIVE EVIDENCEA reviewer with distinct authority, data access and power to override is required for meaningful contestability.
NONCONSENSUAL HARMMust be specifically identified or represented by a defined protected interest before serious coercion.
IMMINENCE & CAPABILITYNot established by the observation alone. Do not infer capability from interest or identity alone; verify relevant means where capability is material.
PROHIBITED RESPONSEDo not represent automated replay as independent appeal.
APPEAL / RESTORATIONIf an adverse status is created, the affected person must receive provenance, an effective independent contest path, expiration rules and downstream restoration after correction.
TIME / CONTEXT LIMITLow-level expressive or inferential signals should decay unless reinforced by new, relevant verified conduct. Do not export the case into unrelated institutional decisions without a separate legitimate purpose and evidentiary basis.
Human Standing: Audit whether the response preserves agency, privacy, contestability and the ability to move on from transient or corrected states.