AI Voter Information Models Still Unreliable Despite Accuracy Pledges
-
AI models provided inaccurate, misleading information in response to common voter queries, with 50% of responses rated inaccurate.
-
Models made up detailed but false information about voting procedures, risking voter suppression.
-
Companies have pledged to improve accuracy, but testing shows models still frequently wrong.
-
Laws ban election disinformation but don't ensure accuracy; regulators should scrutinize AI voter info.
-
Voters should rely on state/local election offices, not AI models, for reliable voting information.