# Magnús Smári Smárason > AI Governance Researcher · Systems Builder · Advisor > Last updated: 2026-07-08. The machine-readable layer (full HTML, JSON-LD, /llms.txt, /llms-full.txt, /sitemap.xml, /robots.txt, /knowledge-graph.jsonld) is maintained and truthful. Personal portfolio and publishing platform. Bilingual (EN/IS). Zero tracking. Based in Akureyri, Iceland. Identity: Wikidata https://www.wikidata.org/wiki/Q140460090 · ORCID https://orcid.org/0009-0008-2050-021X · GitHub https://github.com/Magnussmari · LinkedIn https://linkedin.com/in/magnussmari How to cite: https://smarason.is/en/citation ## About Magnús Smári Smárason is an AI governance researcher and systems builder with a 14-year intellectual arc from surveillance law (2012) through emergency services (2007–2022) to AI governance (2022–present). Author of eight original named concepts in AI governance. Builder of six production systems. MA in Social Sciences (AI and Society), University of Akureyri, grade 9.5/10. Current roles: AI Project Manager at the University of Akureyri, Instructor at Oxford Lifelong Learning, Expert Panel Member at Rannís (Icelandic Centre for Research). ## Key Concepts (Original) - **Responsibility Fog**: Accountability dissolution when decisions distribute across algorithms, committees, and platforms. - **Cognitive Debt**: Judgment atrophy from sustained algorithmic dependence; compounds like technical debt. - **VALOR Framework**: Life-Value governance alternative to EU AI Act risk tiers and NIST RMF. - **Investment-Sentiment Gap**: 41% of AI VC targets tasks workers resist automating; 1.26% targets tasks they want automated. - **Benevolent Cage**: Protective governance that becomes the constraint it was designed to prevent. - **Diagnostic Sociology**: Emergency medicine pattern recognition applied to societal AI effects. - **Cognitive Complacency**: Wealthy societies reduce AI to convenience rather than capability amplification. - **Access Shock**: AI disrupts the access economy of expert knowledge, not just labor markets. All concepts grounded in John McMurtry's Life-Value Onto-Axiology. ## Publications - [Beyond Fragmentation: A Life-Value Alternative for AI Governance](https://skemman.is/handle/1946/52173) — MA Thesis, 2026, grade 9.5/10 - The Augmentation Divide: Cognitive Debt Stratification in the Generative AI Era — SSRN preprint, 2026 - Trading for Conservation: AI and Arctic Species (with Tom Barry) — preprint, 2025 - The Irreducible Human (with Kristian Guttesen) — forthcoming, Brill Publishers - [Lögbundin gagnageymd fjarskiptafyrirtækja](https://skemman.is/handle/1946/21987) — BA Thesis, 2015 ## Systems - **BORG**: Sovereign hybrid AI governance platform deployed at the University of Akureyri - **Arctic Tracker**: Conservation data platform — 473,102 CITES records for 43 Arctic species ([thearctic.is](https://www.thearctic.is)) - **Gjöll**: Fire fatality database — 113 incidents, 145 deaths, 1968–2026 ([gjoll.is](https://www.gjoll.is)) - **SweepGraph**: Text-to-Neo4j knowledge graph construction ([GitHub](https://github.com/magnussmari/SweepGraph)) - **smarason.is**: This site — Next.js 16, Sanity CMS, zero tracking - **Hús Dagsins**: Interactive map of 716 historic buildings in Akureyri ## Teaching & Courses - **AI Training Grounds / Æfingavellir gervigreindar** ([/en/ai-training-grounds](/en/ai-training-grounds)): Drill-based AI training course. 7 modules, 44 drills, capstone deployment. Beginner to agentic engineer arc. $149–199 one-time. - **AI Training for Beginners / Gervigreindarþjálfun** ([/en/ai-training](/en/ai-training)): Free 60-minute guide. Obsidian-based personal AI infrastructure, 10 parts, zero code. - **Training Projects** (free, beginner-friendly): - PRJ-001: Multi-model research — ocean floor mapping ([/en/prompts/projects/research-deep-ocean](/en/prompts/projects/research-deep-ocean)) - PRJ-002: AI vision — NASA space image analysis ([/en/prompts/projects/vision-space-image](/en/prompts/projects/vision-space-image)) - PRJ-003: Build an app without code ([/en/prompts/projects/coding-build-an-app](/en/prompts/projects/coding-build-an-app)) - PRJ-004: AI Training for Beginners — Obsidian system setup ([/en/ai-training](/en/ai-training)) ## Structured Data - [Knowledge Graph (JSON-LD)](https://smarason.is/knowledge-graph.jsonld): Full structured data — 48 nodes, ~118 relationships, 24 relationship types - [Comprehensive LLM context](https://smarason.is/llms-full.txt): Extended version of this file with complete descriptions - [Sitemap](https://smarason.is/sitemap.xml) ## Contact - Email: magnus@smarason.is - LinkedIn: [linkedin.com/in/magnussmari](https://linkedin.com/in/magnussmari) - GitHub: [github.com/magnussmari](https://github.com/magnussmari)