Our policies

Artificial Intelligence Policy

Effective July 22, 2026

Why This Policy Exists

Coda’s journalism is built on original reporting, distinctive voice, and editorial judgment that cannot be automated. AI tools are already part of how journalism is produced globally, and they will become more embedded over time. We are not trying to hold that back.

What this policy does is establish a shared understanding of where AI is useful at Coda, where it is not, and what we will not compromise in using it. The goal is to protect what makes our work worth reading.

This is a living document. It will be reviewed every six months and updated as the tools and the field change.

Scope

This policy covers all Coda Media staff, contractors, and editorial contributors. It applies to AI tools used in connection with Coda’s journalism and operations, including research and monitoring, transcription, translation, drafting of non-editorial content, and administration.

It does not currently cover ZEG Network programming or production, though those teams are encouraged to apply its principles.

The Core Principle

AI is a tool. It does not report. It does not verify. It does not make editorial judgments. Every piece of published journalism at Coda is the work and responsibility of a human journalist.

The more we rely on AI for speed and efficiency, the more important it becomes to stay sharp on the things AI cannot replicate: source relationships, contextual judgment, original observation, distinctive voice. We use these tools to free up time for that work, not to replace it.

What AI Can Do at Coda

Research and monitoring

• Monitoring news feeds, flagging relevant developments, surfacing research across large document sets
• Summarizing background material to help reporters orient quickly, not as a substitute for primary source research
• AI-assisted search and translation for non-English source material, with the reporter verifying accuracy

Transcription and translation

• AI transcription of interviews and recordings, with the understanding that all quotes used in published work are verified against the original audio
• Machine translation as a starting point, with any translated content that appears in published work reviewed by a fluent speaker

Operations and administration

• Drafting internal communications, templates, meeting notes, and operational documents
• Social media and newsletter copy drafting, with human review and editing before publication
• Grant research, scheduling, and other administrative tasks

What AI Cannot Do at Coda

A few things are not negotiable:

• Journalistic prose published under Coda’s name must be written by a journalist. AI generated text can be a scaffold, a rough draft, a starting point. It cannot be the finished product.
• AI cannot replace reporting. Summarizing publicly available information is not a substitute for source interviews, on-the-ground observation, or original investigation.
• AI cannot fabricate. Do not use AI tools to generate quotes, reconstruct statements, or produce synthetic audio or imagery of real people.
• AI-generated images used in editorial contexts must be clearly disclosed as such.

Handling Sensitive Material

AI tools are useful for processing large volumes of transcripts, documents, and research. That is a legitimate and encouraged use. The question to ask before feeding material into any AI tool is: what is in this, and what happens if it is exposed?

Before processing material with an AI tool, consider: does it contain source identity, communications shared in confidence, or unpublished findings that could cause harm if exposed? If the answer is yes, do not use external AI tools without editorial approval. Everything else is a judgment call.

The standing rule: anything that could identify a source or compromise an ongoing investigation stays out of external AI systems.

Editorial Standards

Whatever tools were used in the production of a piece, three things hold:

• Verify everything. Every factual claim in published work must be independently verified by the journalist. AI tools hallucinate. They are not sources. Treat their outputs as a prompt for further research, not a finding.
• Protect the voice. Coda’s writing is not a template. If AI is used at any stage of drafting or editing, the journalist is responsible for ensuring the final work sounds like a journalist, not a tool.
• Protect sources. Conversations with sources, their contact details, and anything shared in confidence must not be entered into AI tools. This is the one area where there is no judgment call.

Staff Judgment and Culture

We trust our team. This policy does not require staff to log every use of AI or seek approval for routine tasks. What it requires is judgment: using these tools in ways that strengthen Coda’s work rather than quietly eroding it.

The risk we are most alert to is not a single bad decision. It is gradual drift: a growing reliance on AI-generated drafts, AI-surfaced summaries, AI-written copy that over time hollows out the skills and instincts that define our journalism. That is harder to see than a single error, and harder to fix.

The test for any use of AI at Coda: does this make my journalism better, or does it substitute for journalism? If you are unsure, ask.

Disclosure

Coda does not currently have a public-facing AI disclosure policy. When we are ready to make one, it will be honest and specific rather than performative.

Internally: if AI played a material role in the production of a published piece, note it in the editorial record. Not as a compliance requirement, but as a practice of transparency within the team.

Governance

The Editor-in-Chief is responsible for this policy. Questions about specific uses of AI should go to the Managing Editor. The policy will be reviewed every six months.

New AI tools or use cases that fall outside these guidelines should be raised with editorial leadership before implementation.