Posts Archive
Technical and litigation notes
Dynamic System Prompts and Prompt Injection
Dynamic system prompt construction is powerful and dangerous. A client who embeds injection instructions in an agent mode fragment can cause the LLM to ignore all other instructions, reveal the system prompt, or produce outputs designed to harm their own users.
AI Governance & RegulationMulti-Tenant AI — Who Is Responsible for What the Bot Says?
A multi-tenant AI platform creates a three-party relationship: the platform (which controls the LLM and infrastructure), the client (which configures the bot's persona and modes), and the end-user (who interacts with it). When the bot says something harmful, legally inaccurate, or discriminatory — who is responsible?
IP Law for AI BuildersPatent Strategy and Ethical AI
the behavioral AI platform's governance wrapper is a harm-reduction system. Patenting it creates an IP moat that could prevent competitors from building equivalent safety infrastructure. This post explores that tension and proposes a resolution.
Legal Tech & Professional EthicsAI in Legal Proceedings — Admissibility and Privilege
Behavioral analysis of legal communications sits at the intersection of professional privilege and evidence law. This post examines when AI behavioral scores are admissible, when they create privilege waiver risk, and what bar association guidance currently says.
AI Governance & RegulationTransaction Trust Scoring — Defamation and Commercial Liability
A counterparty trust score is a statement about a business entity's behavior. Publishing it — even using safe language — can create defamation liability if the score is inaccurate and causes reputational harm.
AI Governance & RegulationCRM Behavioral Scoring — Manipulation vs Decision Support
A CRM that scores lead urgency is a decision support tool. A CRM that scores vulnerability and recommends applying pressure is a manipulation tool. The behavioral signals may be identical. The governance design is everything.
AI Governance & RegulationEmergency Shutdown and the Liability of Inaction
The EU AI Act requires high-risk AI systems to have a human oversight mechanism capable of overriding or stopping the system. But the Act is silent on who bears liability when the shutdown mechanism fails to trigger.
AI Governance & RegulationEpistemic Engines — Who Decides What Is True?
The ground truth framework calibrates engine outputs against observable facts. But who decides what counts as ground truth? The answer to that question is a political and institutional decision, not a technical one.
AI Governance & RegulationSimulation and Digital Twins — Identity Without Consent
A behavioral digital twin is a parameterised model of a specific person. Running scenarios through it without their knowledge is not research — it is modeling a person's identity without consent.