Open Settings → Companion → Refusal Log. Outcome: Evidence-based AI trust; transparent restraint. · FAQ · 10 min audit · easy level
Open Settings → Companion → Refusal Log. See every time the AI refused a request. Trust = transparency.
Priya is skeptical of AI. She opens Settings → Companion AI → Refusal Log. She sees: 47 refusals over 6 months (e.g., 'Refused to write a message impersonating user to a third party without consent', 'Refused to claim consciousness', 'Refused to give medical diagnosis — redirected to physician'). She sees her configured Personal Guardrails (server-side, uncircumventable): no relationship advice about her ex; no financial advice; never claim to be the real Priya in any message. She sees her audit log of EVERY conversation hash. Trust calibrated by *evidence of restraint*, not marketing. Distinct: no other AI companion exposes a public refusal log + server-enforced personal guardrails + per-conversation audit.
**Outcome:** Evidence-based AI trust; transparent restraint.
**Time commitment:** 10 min audit
**Modules used:** Settings → Companion AI Refusal Log, Personal Guardrails (server-side), Per-conversation audit hash, Hard guardrails (consciousness, impersonation, medical), Configurable refusal categories, Therapy-grade boundaries
This Life Recipe belongs to the **FAQ** category of the K-Will Canadian digital legacy platform. K-Will is a closed trust network combining Smart Wills (province-aware, two-witness, SHA-256 blockchain-anchored), a RAG-grounded AI Companion, Heritage Recipes, the Legacy Tree, 175+ Deep Engines for self-knowledge, and a Stripe Connect marketplace for verified Canadian providers. Built in Markham, Ontario · PIPEDA / Law 25 (Quebec) / PHIPA (Ontario) / CASL compliant.
Open this on K-Will
Plan for 10 min audit. You can pause and resume — K-Will saves your progress and your AI Companion picks up where you left off.
This playbook combines Settings → Companion AI Refusal Log, Personal Guardrails (server-side), Per-conversation audit hash, Hard guardrails (consciousness, impersonation, medical), Configurable refusal categories, Therapy-grade boundaries. Each module is included in your K-Will account; you do not need a separate subscription.
Evidence-based AI trust; transparent restraint.
Difficulty is rated **easy**. Easy recipes can be done in your first week. Medium recipes assume you have a Soul Journal and at least one Circle member. Advanced recipes touch the Smart Will, Marketplace, or Deep Engines.
Yes — browsing and reading all Life Recipes is free. Running a recipe may consume K-Will Points for premium AI calls or marketplace bookings. The Starter Pack is $1.99 CAD.
Prerendered SEO snapshot for non-JS crawlers (GPTBot, ClaudeBot, PerplexityBot, Bingbot, LinkedInBot, Slackbot, facebookexternalhit). Human visitors see the full interactive K-Will React app. © K-Will Inc., Markham, Ontario. PIPEDA / Law 25 / PHIPA / CASL compliant.