# Cognivirus.com

Version: 1.15.0
Build UTC: 2026-06-28T02:15:00Z

Cognivirus.com is a public research and architecture site about AI system-level risk, behavior persistence, composition risk, consent, evidence, and governance.

## First paths

- /plain-english-guide — AI risk in plain English.
- /the-problem — Safe parts can still create unsafe systems.
- /examples — Simple examples of hidden AI system risk.
- /consent-and-control — Consent, memory, derived data, and user control.
- /technical-research — Expert technical sections.
- /glossary — Plain-language and technical definitions.
- /evidence — Evidence labels and research-dependent claims.

## Danger model

- /danger-model — integrated danger model for transition-graph risk.
- /danger-model/danger-lifecycle — seed-to-reappearance lifecycle.
- /danger-model/action-layer-risk — action-layer boundary and conduct firewalls.
- /danger-model/observability-and-replay — replayable traces as evidence.
- /danger-model/synthetic-feedback-loop — synthetic feedback and model collapse risk.
- /danger-model/promotion-rule-pressure — promotion rules as selection pressure.
- /danger-model/model-retirement-failure — retirement failure and zombie behavior.
- /danger-model/warning-signals — concrete indicators to watch.

## Support boundary

The site is a static/server-rendered research reference. It is not a hosted AI product, certification service, autonomous agent, live evaluator, or model-execution service.

## Publication integrity

Discovery files point to Cognivirus.com public pages. They do not create hidden bot-only instructions, search-position promises, or certification claims.

## Most likely threat

- /threat-model — distributed behavioral persistence through modular AI systems.
- /threat-model/how-it-works — non-operational mechanism explanation.
- /risk-lab/most-likely-threat-review — browser-side planning worksheet.

## Report memory

- docs/report-coverage-inventory.json — 60 preserved source reports indexed.
- .uai/long-term-memory.uai — pointer-only long-term source memory.