AI Disclosure

1. Overview

ChamberLight uses artificial intelligence to assist with content analysis and make information more accessible. This page explains exactly how AI is used on the Platform, what model powers it, what AI does not do, and the limitations you should be aware of.

2. How AI Is Used

ChamberLight employs AI for the following specific purposes:

• Article summarization: When a news article is submitted, AI generates a concise summary to help users quickly understand the key points without leaving the Platform.
• Source credibility scoring: AI evaluates the credibility of the source publication based on factors such as editorial standards, correction policies, and historical accuracy. A source's score combines these ratings over at least 10 of its stories; see How source scores work. This score is displayed alongside submissions as additional context.
• Article neutrality assessment: AI analyzes individual articles for political lean on a left-right spectrum, evaluating factors such as language tone, source selection, framing, and headline slant. The neutrality score ranges from 0 (far-left) to 100 (far-right); 50 means no detectable left or right tilt. Lean is not balance: how even-handed an article is gets a separate rating, from 0 (entirely one-sided) to 100 (presents the sides fairly). A source's lean combines the lean ratings of at least 10 of its stories; see How source scores work.
• Bill plain-English translation: When legislative bills are referenced, AI generates plain-English explanations of what the bill does, who it affects, and its key provisions, making complex legislative text accessible to all users.
• Votes and amendments: AI writes plain-English explanations of congressional roll-call votes and amendments. Vote tallies, thresholds, party breakdowns and the “commonly misread” warnings shown beside them are computed from official records, not by AI.
• Candidate profiles: AI drafts short candidate biographies and issue summaries only from the public sources it is given (such as Wikipedia, Wikidata and the candidate's own website), with links to those sources.
• News coverage grouping: AI helps group articles about the same event, writes neutral summaries of how different outlets framed a story, and classifies news sources by type (for example, wire service, local news or government).

3. What Models We Use

ChamberLight's AI features are powered primarily by GPT-6 Luna, a large language model developed by OpenAI (opens in new tab), accessed through OpenAI's API. If that service is unavailable or does not return a usable answer in time, the same request is sent instead to Gemini 3.8 Flash, a large language model developed by Google (opens in new tab), accessed through Google's Gemini API. The model that produced each analysis is recorded internally.

No personally identifiable user information is sent to either provider—only the text of articles, bills, amendments and votes, publicly available information about candidates and officials, and the source metadata needed for analysis.

4. What AI Does NOT Do

It is important to understand the boundaries of AI on ChamberLight. AI does not:

• Make editorial decisions: AI never decides what content is published, featured, or promoted. All editorial and publishing decisions are made by human moderators and the community.
• Determine categories: The Praised/Mixed/Called Out categorization of submissions is based on community input and moderator review, not AI classification.
• Moderate content: AI does not approve, reject, remove, or quarantine submissions. All moderation actions are performed by human moderators following our Moderation Policy.
• Generate or author content: AI does not write submissions, comments, or any content that is attributed to users or presented as original reporting.

5. Versioning & Reprocessing

As AI models improve, we may reprocess previously analyzed content with updated models to provide more accurate summaries, credibility scores, and neutrality ratings. When reprocessing occurs:

• The AI-generated content will be updated to reflect the improved analysis.
• The model version used for each analysis is recorded internally for audit purposes.
• Significant changes in methodology or scoring will be announced to users.

6. Limitations

AI is a powerful tool but it is not infallible. Users should be aware of the following limitations:

• AI can make mistakes: Summaries may omit important nuances. Credibility and neutrality scores are algorithmic assessments, not definitive truths.
• Scores are not endorsements: A high credibility score does not mean ChamberLight endorses the source or its reporting. A low score does not mean the content is false.
• Context matters: AI analyses are based on patterns in text and may not capture the full context of a political situation, regional significance, or emerging story.
• Not a substitute for reading: AI summaries are provided as a convenience. We encourage users to read the original source article for a complete understanding.

7. Contact

If you have questions about how AI is used on ChamberLight, or if you believe an AI-generated analysis is inaccurate, please contact us at ai@chamberlight.com.

To challenge a source's score or label, use “Challenge this score” on the source's page. An editor reviews every challenge, and the outcome is published on that page; see Challenging a score.