How the Oversight Wall works, and what its limits are

What is on the Wall

Each card on the Wall is an "oversight gap" — a place where this platform's documented public-record evidence (court filings, regulatory settlements, financial disclosures, voting records) suggests harm is on the record but accountability is missing. We do not generate evidence; we surface patterns in evidence already on file.

The six gap types

• Enforcement gap — an entity has multiple documented violation facts and no documented connection to a regulator or prosecutor.
• Pattern gap — multiple separate entities share the same enabling third party and each independently has documented violation history. The classic case is a bank whose multiple major clients all turn out to be running fraud.
• Disclosure gap — an elected official has accepted documented donations from a sector with material interests in a specific topic, and this platform's records contain no public statement, vote, or hearing by them on it. This is a claim about our records, not the world: a statement can exist that we never ingested, which is why every disclosure gap invites a correction.
• Recusal gap — an official is part of a documented closed loop of authority and benefit, with no documented recusal from related decisions.
• Jurisdictional gap — a non-US entity has documented cross-border ties to US-tracked entities but no documented US regulatory filing or enforcement action.
• Legislative gap — a sector has accumulated documented violations across multiple companies in a recent window with no documented congressional hearing in the same window.

Confidence and severity

Every gap inherits the lowest confidence of its supporting evidence. We use only three levels:

• primary — supported by direct primary-source evidence (court filing, agency press release, official disclosure)
• secondary — supported by reputable reporting that itself cites primary sources
• inferential — derived from a documented pattern, not from any single direct citation. Inferential gaps are the most fragile and are labelled as such throughout the site.

A gap with no attached evidence rows is always labelled inferential — it cannot borrow confidence it does not carry. Silence-based gaps (disclosure gaps) are negative-evidence detections and are capped at secondary no matter how strong their attached evidence is: the donations are documented, but the silence itself is an absence in our records, not a citation.

Severity: a gap supported by at most one piece of attached evidence is always low — no other factor can raise it. Past that floor, severity is high when there is ≥ $100M of documented impact, ≥ 5 affected entities, or the most recent supporting evidence is more than 2 years old; medium otherwise.

What the Wall does not do

• It does not allege individual wrongdoing. A gap is a question, not an accusation.
• It does not act on a reader's behalf. We never send FOIAs, emails, or comments for you. Templates are copy-paste tools.
• It is not exhaustive. Absence of a gap on the Wall is not evidence of accountability.
• It does not use AI to generate civic-action templates. Every template is hand-curated for legal reliability.

Corrections

If you are the subject of a gap, or have evidence that a gap has been resolved or that supporting evidence is wrong, please use our corrections process. We attach responses to the relevant page.

Freshness and retirement

Detection runs nightly. Each gap shows the date its supporting evidence was most recently verified. Gaps whose evidence has not been refreshed in 90 days are marked as stale.

A gap that a nightly run no longer derives is automatically retired from the Wall — usually because the accountability action it flagged as missing (an enforcement action, a public statement, a recusal, a hearing) has since been documented. Retired gap pages stay reachable, marked as retired, for the record; if later data re-derives the gap it returns to the Wall. A detector that fails to run leaves its gaps untouched rather than mass-retiring them.