Vote AI Artificial Intelligence Disclosure

1. Purpose & Scope

The purpose of this AI Disclosure statement is to inform citizens, candidates, journalists, and election officials exactly how Vote AI utilizes artificial intelligence. We believe voters have an absolute right to know when they are interacting with AI, how answers are derived, and what technical guarantees prevent bias and misinformation.

2. How AI Personas Generate Answers

Candidate personas on Vote AI are powered by generative artificial intelligence models (such as Google Gemini 2.5 Flash) operating strictly within a Retrieval-Augmented Generation (RAG) architecture. When a user asks a question:

• The question is embedded as a vector and matched against candidate-specific documents stored in Pinecone vector databases.
• Relevant excerpts from verified campaign documents, speech transcripts, legislative votes, and policy papers are retrieved.
• The generative model synthesizes a natural-language response using only the retrieved text excerpts.
• If the candidate has not taken an official stance on the issue, the model is instructed to refuse to speculate and explicitly clarify that no record exists.

3. Source Citations & Anti-Hallucination Controls

Unlike open-ended consumer conversational models that fabricate plausible facts when memory gaps occur, Vote AI employs strict anti-hallucination guardrails:

• Every statement outlining a policy position is accompanied by an interactive citation badge.
• Voters can click the citation badge to inspect the original underlying speech, bill number, or whitepaper.
• Candidate personas are forbidden from editorializing or adopting political opinions not present in their verified records.

4. Algorithmic Neutrality & Nonpartisanship

Vote AI is strictly nonpartisan. We do not endorse candidates, political parties, ballot measures, or political action committees. Our software architecture applies identical processing pipelines, vector search algorithms, and response generation prompts to all registered candidates, regardless of party affiliation, ideology, or campaign budget.

5. Voice & Likeness Protection Standards

We hold synthetic media to the highest ethical benchmarks:

• Voice mode is only activated with the candidate’s verified cryptographic consent.
• All audio synthesized by Vote AI contains inaudible digital watermarks (C2PA/SynthID) identifying it as synthetic speech.
• Unverified third parties are strictly blocked from generating candidate voice or likeness media.

6. VoteBot Neutral Civic Guidelines

VoteBot is our independent civic guide designed to help citizens understand the voting process, register to vote, locate polling places, and explore ballot measures. VoteBot is programmatically constrained to neutral explanations of ballot initiatives, summarizing both official proponent and opponent arguments with equal weight.

7. Auditing & Reporting Hallucinations

If you observe a response that appears inaccurate, biased, or misattributes a position, every message includes a report button that submits the transcript to our AI Safety team. Inaccuracy reports are investigated within 24 hours.

Direct policy audit requests: [email protected]