MICA: an AI-alignment concept for higher stakes
MICA is a set of instruments that read what an AI agent does and print what they can and cannot see. Its method is published as the MICA canon, 50 chapter files, version 0.1.0. This paper introduces the canon. It tells the architecture as one account: the two questions MICA separates, the health substrate, the four gates, the seats around a deployment, the principles, and what the canon holds over the whole chain. It states once what is designed, what is built and what has run, and it shows how to read the canon, alone or with an AI assistant. Every statement cites the chapter and section it rests on, and an annex lists all 50 chapters. The file mica-canon-v0.1.0.zip is the exact archive of tag v0.1.0 of https://github.com/arcencilo-family/MICA-canon. Drafted with Claude (Anthropic) and audited against the canon. The claim tags follow The Science of Not-Yet-Knowing (https://doi.org/10.5281/zenodo.23961975).
Authors
- Frank Böhm
Publication Details
- Journal
- Zenodo (CERN European Organization for Nuclear Research)
- Published
- 2026-11-17
- DOI
- https://doi.org/10.5281/zenodo.24322044
- Primary Topic
- Ethics and Social Impacts of AI
- Type
- preprint