In plain English
This page explains where an AI behavior can live. It may be in a model, but it may also be in a prompt, memory record, adapter, dataset, tool setting, evaluator rule, or human workflow.
- Why this matters: AI risk can come from the whole arrangement, not one obvious model.
- What to look for: data, memory, routes, adapters, tools, evaluators, updates, and rollback paths.
- Technical version below: the expert terminology remains available and is linked through the glossary.
Cognitive and Human Hosts
A cognitive hostA part of an AI system that can carry or express a behavior. Open glossary definition is anything capable of carrying or expressing a pattern. The reports expand this beyond artifacts to include workflows, incentives, and human procedures.
Artifact hosts
Artifact hosts include base models, LoRA adaptersA common kind of small adapter used to specialize large models. Open glossary definition, prompt packages, policy files, memory stores, synthetic datasets, routing rules, evaluator prompts, tool manifests, registry aliases, and release notes.
Process hosts
Process hosts include approval habits, incident-response runbooks, dashboards, metric thresholds, canary rituals, hidden-test custody, and escalation norms. These hosts can preserve behavior by making certain changes easy and other changes difficult.
Human hosts as a safety term
The site uses “human host” only as an analytical term for incentive exposure. It does not blame users or operators. It asks how systems can preserve human agency, exit, and dissent when AI tools become useful, intimate, or status-enhancing.