PSYBORG / EVIDENCE METHODOLOGY

Show the basis.
Keep the limits.

A transparent reading guide to a small, connected collection. Version 1 · 8 October 2026.

What is in the collection?

Public entries require a source and a review date. Risks describe pathways to harm; incidents describe documented or attributed historical events; research records describe evidence. Models are version-specific. Regulations retain jurisdiction and applicability caveats.

The starter collection is historical and deliberately small. It does not represent incident prevalence, comparative model safety or an exhaustive watchlist. Records are maintained by the publisher; there is no automated ingestion or monitoring service behind this release.

Verification is not publication status.

Confirmed: Occurrence supported by direct documentation, a responsible authority’s finding or an originating organization’s acknowledgment. This does not establish every claimed consequence.
Reported: A sourced account exists, but occurrence has not been independently established to the same standard in this record.
Allegation: A claim in a complaint or from an identifiable party. Attribution stays visible; publication does not turn it into a finding.
Scenario: A possible future or hypothetical sequence. Scenarios belong in research, never the historical incident database.

Confidence and severity.

V1 assigns no probability, confidence or severity scores. A universal rank would conceal differences in evidence, context and affected populations. Each risk instead describes what is known, what is unknown and the key uncertainties.

How connections work.

A relationship is stored once and shown in both directions. Labels explain whether a record develops, involves, examines or addresses another. A regulator can be involved as an authority; that does not mean it caused harm. A regulation-to-risk link is topical context, not a finding of legal applicability.

Freshness and corrections.

Event dates, publication dates and review dates are separate. Unknown dates stay unknown; year-only and month-only sources are not converted to invented days. Records reviewed more than six months ago display a review-due notice. Editors supply a public change note with each revision. Private notes and draft revisions remain private.

Evidence classifications.

Labels describe the kind of evidence or contribution. They do not certify correctness. A regulator complaint can be a primary source while its substantive claims remain allegations.

Primary source

Direct evidence from a participant, original record or originating institution; not an independent truth guarantee.

Official documentation

Documentation issued by the responsible organization. Claims may be self-reported.

Government publication

A public authority publication. Allegations, findings and guidance remain distinct.

Peer-reviewed research

Research whose peer-review status is documented in the citation. Review does not eliminate uncertainty.

Institutional research

Research issued by an institution; peer review is not implied.

Technical report

A technical report or evaluation; results apply to its stated systems and conditions.

Established journalism

Reporting with named authors and an identifiable editorial publisher; consult the original evidence where possible.

PSYBORG analysis

PSYBORG interpretation of cited material, distinguished from the underlying evidence.

Community research

Community-produced evidence whose methods and limitations require scrutiny.

Opinion

An attributed judgment, not an empirical finding.

Forecast

An expectation about the future, not an observed event.

Scenario

A possible sequence of events; not an incident that is known to have happened.

Hypothesis

A testable proposed explanation that remains to be established.

Speculation

A possibility with limited support; not an established conclusion.

Different timelines, different questions.

Intelligence Timeline contains documented historical developments. AI Futures contains speculative capability forecasts and attributed opinions. Neither is a countdown to catastrophe.