In plain English
This page explains the governance layer: rules, logs, approvals, signatures, audits, permissions, and rollback tools. These controls are necessary, but they also become important failure points.
- 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.
Source Report Intake Governance
The uploaded reports were treated as active source intake. Each report was copied to a local durable source path, summarized into a public-safe source-dossier record, added to the evidence manifest, hashed in the integrity manifest, and recorded in .uai/intake-outcome-ledger.uai.
Why raw reports are not the public page
Some reports use speculative or adversarial scenario language. Preserving them is useful for provenanceA record of where a component or behavior came from. Open glossary definition and future review. Publishing their operational details as public guidance would conflict with the site’s safety boundary. The public site therefore uses a layered model:
- Raw source report preserved locally.
- Public source summary with evidence label and limitations.
- Research index with source-dossier status.
- Public guide pages that paraphrase safe analytical content.
.uaimemory that records what moved where.
Disposition values
Every uploaded report receives one or more dispositions: incorporated into public content, preserved to durable memory, or retained as raw source material. If a future file is rejected, the ledger must record the reason.
Required future behavior
Future agents should scan agent-file-handoff/Content, agent-file-handoff/Improvement, and .uai/intake-outcome-ledger.uai before claiming completion. A file handoff is incomplete while active intake lacks disposition, processed outcome, and proof of use.