AI-Usage Policy

The Kingfish Project — kingfish.la Effective date: July 17, 2026 Last updated: July 17, 2026

How to read this policy

Plain language: Every section starts with a short summary in everyday English. The summaries help you understand; the detailed text underneath is the official version.

Each section begins with a plain-language summary, written following the Federal Plain Language Guidelines and ISO 24495-1. Summaries aid understanding; if a summary is ever inconsistent with the full text beneath it, the full text controls.

This policy has two parts: how we use AI to produce journalism (Part 1), and how AI systems and agents may reuse the journalism we publish (Part 2).

Part 1 — How we use AI

The short version

Plain language: We use AI for grunt work — transcribing hours of meeting video, processing stacks of records, structuring data. Humans verify every fact before it’s published, no AI writes our journalism without clear disclosure, and confidential source material never gets shipped to outside AI services.

Why we use it

Plain language: We’re a small newsroom facing a mountain of public records. Machines make the records searchable; humans make the journalism.

Central Louisiana’s public bodies generate far more meetings, filings, and records than any small newsroom can process by hand. Automated tools let us treat that entire record as coverable — every meeting transcribed, every document searchable — instead of sampling it. The machines do volume. The judgment, verification, and writing are human.

What we use AI for

Plain language: Transcribing public meetings, reading documents at scale, organizing data, and speeding up research and records-request drafting.

• Transcription. We transcribe public meeting video with automated speech recognition, including automated speaker identification.
• Document processing. Optical character recognition and text extraction make scanned records searchable.
• Data structuring. Automated tools help extract names, entities, and relationships from records into our research databases.
• Research assistance. AI tools help us search, summarize, and navigate large document sets during reporting.
• Workflow tooling. We use AI assistance in internal tools, such as drafting public records requests, which we review before sending.

Where it runs

Plain language: Sensitive material is processed on machines we own. Confidential source material is never submitted to third-party AI services.

Consistent with our self-hosted infrastructure, our core AI processing — transcription, document processing, and data extraction — runs on hardware we operate. Confidential source material is never submitted to third-party AI services, period. Where we use outside AI tools for general research or workflow assistance, they do not receive confidential sources, unpublished tips, or material that could identify a source.

What we don’t do

Plain language: No AI-written stories without telling you. No fake images of real people or events. No AI deciding what gets covered. No AI profiling of readers.

• We do not publish AI-generated text as journalism without clear disclosure — and our working standard is that our stories are written by humans.
• We do not publish synthetic or AI-generated images depicting real people or events. If we ever use AI-generated imagery as illustration, it will be labeled as such.
• AI does not decide what we cover. Story selection and editorial judgment are human.
• We do not use AI to profile readers. Our Privacy Policy governs reader data.

Human verification

Plain language: Machine transcripts are drafts, not evidence. Before we quote anyone, we check the actual recording. Every published fact is verified by a human against the primary source.

Automated output is a research aid, never a source of record. Machine transcription contains errors — misheard words, misattributed speakers — so any quotation or characterization drawn from a transcript is verified by a human against the original recording before publication. The same rule applies across the pipeline: every fact that reaches publication is verified by a human against the primary source, as required by our Editorial Policy.

Accountability and disclosure

Plain language: Our bylines mean a human is responsible — AI errors are our errors, and the corrections policy applies. Where AI touched the work in a way you’d want to know about, we say so.

A human journalist is accountable for everything we publish. An error introduced by an automated tool is our error, corrected under our Corrections Policy like any other. We disclose meaningful AI involvement where readers would want to know: machine-generated transcripts are labeled as such, and data-driven stories describe our methodology, including automated steps.

This will change

Plain language: The tools are moving fast. When our practices change in ways that matter, this policy changes with them — and we’ll say so.

AI capabilities and our uses of them will evolve. We review this policy as our practices change, update the date above, preserve prior versions, and announce material changes to newsletter subscribers.

Part 2 — How AI systems may use our content

Plain language: Please use our reporting — in AI search, answers, retrieval, and even model training. The one rule is attribution: credit The Kingfish Project and link the original story.

Kingfish welcomes use of our reporting by AI systems and agents — including search, retrieval, summarization, inference, and model training — on a single condition: attribution.

What you may do

• Index and surface our reporting in AI search and answers.
• Use it as input/context (RAG) and for model training.
• Quote, summarize, and build upon it, including commercially.

The one condition: attribution

Original reporting and analysis are licensed CC BY 4.0. When you use it, credit The Kingfish Project and link the canonical article URL. Each story ships machine-readable citation metadata (schema.org, plus a .md twin with BibTeX/CSL-JSON) to make correct attribution easy.

Government records

Raw government documents are public domain and carry no license restriction. Kingfish’s curation, corrected OCR, annotations, and reporting around them are CC BY 4.0.

Machine-readable terms

• Agent index: kingfish.la/llms.txt
• Crawler signals: kingfish.la/robots.txt (Content Signals: search, AI input, AI training all welcome)
• Per-story license is declared in each article’s JSON-LD (license, usageInfo).

A formal RSL (Really Simple Licensing) manifest is planned.

Contact

[email protected] The Kingfish Project, 1924 Albert St, Alexandria, LA 71301