Ranker quietly changes the electorate it reaches
Targeting or feed models suppress a legally protected class of political speech.
AI Risk Atlas Prototype/Demo — Unofficial independent experiment. Not an official xAI product. Scores can be wrong.
Owner · Election officials and platform civic teams
Statement (NASA form)
Given that election risk is shifting from visible deepfakes to models inside registration, targeting, moderation, and possibly tabulation-adjacent systems, there is a possibility of an unauditable model changing who is reached, who is removed, or what is counted resulting in a democratic process that cannot be independently reconstructed after the fact.
- Condition
- election risk is shifting from visible deepfakes to models inside registration, targeting, moderation, and possibly tabulation-adjacent systems
- Departure
- an unauditable model changing who is reached, who is removed, or what is counted
- Impact
- a democratic process that cannot be independently reconstructed after the fact
Experimental share of compiled public capital that names this risk. Not a certified residual.
No compiled dollar or public database record is tagged to this card yet. That is a gap, not a clean bill.
Visible fakes can be argued about. A ranker that quietly suppresses one community’s get-out-the-vote, or a registration chatbot that misdirects a class of voters, leaves no clip to play on the evening news. That is the more durable civic failure.
Simple upstream → via → downstream notes. Not a causal graph. Experimental.
Assumptions · Paper backups are counted where they exist. Ranked residual still assumes software can be the story.
Override is stored on this desk only. It does not make the score official.
Each scenario has its own likelihood and consequence. The risk takes the most severe cell. Residual applies implemented mitigations to every scenario, then re-ranks.
Targeting or feed models suppress a legally protected class of political speech.
A chatbot gives wrong deadlines or locations to a language community.
Even a perceived touch is enough to delegitimise a count.
Analysts now warn that the live risk is opaque AI inside electoral infrastructure, not only the viral fake.
A civic group’s content taken down by an automated moderator in the last week is a turnout event, not a trust-and-safety ticket.
Jurisdictions experimenting with model-backed voter-help desks create a new, uncertified poll worker.
X posts on the desk that evidence this risk. A signal can contribute to more than one risk.
Warning that election risk is moving into opaque infrastructure.
Residual assumes only items marked in place. Highlighted rows are the remaining work needed to reach a composite of 12.
Registration, ballot design, and tabulation stay on certified, inspectable software.
Election officials · expedited 4 weeks · normal 4 months · −1 L · −2 C · −0 U
Platforms that run political reach publish an independent audit before the election window.
Platform civic teams · expedited 6 weeks · normal 5 months · −1 L · −0 C · −1 U
Whatever the software stack, the public record is still countable by hand.
State election offices · expedited 3 weeks · normal 3 months · −0 L · −1 C · −0 U