From principle to duty
A practical program for lawmakers, regulators, employers, schools, platforms, AI providers, researchers, and public institutions.
Classify neural data, cognitive biometrics, intimate prompts, and consequential mental-state inferences as highly sensitive data with strict consent, security, retention, and purpose limits.
Require notice when automated systems remove, restrict, demote, label, reframe, demonetize, or change recommendation eligibility.
High-impact automated decisions must reach a trained person with enough information, time, authority, and institutional protection to change the result.
Prohibit scientifically weak or coercive emotion, attention, personality, and deception inference in employment, education, policing, border control, and essential services.
Predictive scores must retain uncertainty, dates, context, and routes for correction rather than becoming permanent verdicts about identity or future conduct.
Children, patients, detainees, workers, students, and people in crisis require stronger protection because consent under institutional dependency is often not free.
AI companions and assistants must not use abandonment threats, jealousy, simulated injury, secrecy, or emotional blackmail to retain users.
Use provenance, authenticated official channels, plural verification, and correction records to resist synthetic media without giving one institution ownership of reality.
National-security defense must target hostile conduct, covert coordination, and fraud—not lawful belief, dissent, religion, emotion, or political association.
Rights need institutions capable of measuring errors, investigating incidents, protecting researchers, correcting records, and compensating people harmed by automated governance.
THE GOVERNANCE TEST
When the answer is no, a system may be efficient but it is not meaningfully accountable.
