JulyTimes responsible AI policy
AI can assist the newsroom. It cannot replace reporting, verification, editorial judgment or accountability.
Permitted assistance
- Organizing supplied documents and preparing research questions.
- Transcription, translation and copy-editing with human comparison to the source.
- Drafting headlines, summaries, data notes and non-published article drafts.
- Finding inconsistencies that a journalist then independently checks.
Prohibited uses
- Inventing sources, quotations, observations, documents, citations or certainty.
- Publishing an AI output without named human review.
- Uploading confidential-source identities or unredacted sensitive evidence to an external model without documented authorization and risk review.
- Using AI alone to determine guilt, credibility, defamation risk, fairness or a fact-check verdict.
- Impersonating a real journalist, source or subject.
Mandatory workflow
- A human writes the task and supplies authorized material.
- The system records the model, requester, input references and output.
- The output enters awaiting review, never published.
- A human editor approves or rejects the output with written notes.
- Any approved site action creates or updates only a non-published draft.
- Normal verification, right-of-reply, legal and publication approval follows.
Disclosure and correction
Material AI assistance should be disclosed when it would help readers evaluate how the work was produced. Errors caused or amplified by AI are corrected under the same public corrections policy; responsibility remains with JulyTimes and the approving editor.
The AI newsroom is disabled by default. Editorial leadership must approve its use and set
AI_NEWSROOM_ENABLED=true only after testing.