Posts Archive
Technical and litigation notes
Memory Engines and the Right to Be Forgotten
A behavioral knowledge graph stores facts about a person derived from their interactions. GDPR gives that person the right to erasure. But deleting a fact from a graph without deleting the facts derived from it is not erasure — it is redaction.
AI Governance & RegulationEmotional and Psychological Engines — Diagnosis Without a License
Emotional regulation engines score behavioral indicators. The moment those scores use clinical language, the developer has crossed into diagnosis — an activity requiring a license.
AI Governance & RegulationBehavioral Prediction Engines and the Presumption of Innocence
Long-horizon behavioral prediction engines produce 90-day trajectories. This post examines when prediction becomes pre-judgment, and what legal protections should apply.
AI Governance & RegulationOntology as Power — Who Defines Behavioral Reality?
A behavioral ontology decides what concepts exist, how they are defined, and which cultures they assume. The team that writes the ontology defines the system's model of human behaviour. That is a political act as much as a technical one.
AI Governance & RegulationGoverning AI Safe Language — Who Writes the Rules?
The governance wrapper translates harmful labels into safe phrases. But who decides which phrases are safe? This post examines the governance deficit in safe-language design.
Data Privacy & GDPREvent Buses and Data Minimisation
A behavioral event bus logs every signal emitted between engines. Those signals are personal data. This post examines the conflict between event logging for audit and GDPR data minimisation.
AI Governance & RegulationExplainability as Legal Evidence
An explainability trace generated at scoring time can be produced in discovery. This post examines chain-of-custody requirements, forensic soundness standards, and the gap between technical explainability and legal admissibility.
AI Governance & RegulationConfidence Scores and the Illusion of Precision
A confidence score presented without its uncertainty band and reducer list becomes a claim of certainty. This post explores the liability of precision theatre in AI outputs.
AI Governance & RegulationThe Control Plane as a Governance Chokepoint
The control plane knows everything: which engines ran, in what order, and with what outputs. That makes it the ideal — and perhaps the mandatory — place to enforce jurisdiction-specific AI regulation.