How we classify positions

One measure at a time

A position is always about one specific measure: a bill, amendment, or other concrete proposal. We do not publish a universal pro-AI or anti-AI score.

Positions are recorded for specific measures based on dated votes, bill sponsorship, and verified public statements. We do not assign an overall pro- or anti-AI score to anyone.

The measure

We first classify the measure itself: does it strengthen oversight, weaken it, mix the two, or leave the effect unclear? That classification has a written rationale and is approved by an editor.

The evidence

An official's position on that measure uses dated evidence, in this order: a recorded vote, sponsorship or cosponsorship, then a public statement a human has checked against the source.

Supports means the evidence shows the official backed that measure. Opposes means the evidence shows they voted or spoke against it. Mixed means the evidence conflicts. Insufficient evidence means we found none of those records.

“Insufficient evidence” means we found no recorded vote, sponsorship, or verified statement from this official on this measure. It is not a judgment about their views.

Approval

Every approved position shows its evidence links, the methodology version, and the date it was last verified. Agents may draft a position. Only an editor can approve one. A second AI review is not a guarantee of accuracy.

Oversight is an independent service. It is not a government app.