Summary

CJR reports on JESS, a beta chatbot built to help journalists plan for physical safety, digital security, legal risk, and assignment-specific precautions. The story matters because it shows a practical, bounded newsroom use case for AI that is neither content generation nor generic productivity: a source-linked safety-planning assistant for freelancers and smaller outlets that cannot afford dedicated trainers.

Why It Matters

This is a strong direct journalists story because it documents AI being operationalized inside a concrete reporting-support workflow.

  • It shows AI being used for assignment planning rather than writing or rewriting journalism.
  • It names the specific risks the system is meant to cover, including protests, storms, online harassment, and legal exposure.
  • It surfaces the implementation choices that make the tool more defensible: a closed corpus, source links, tester feedback, and explicit privacy limits.
  • It keeps the human role central by treating the system as a gap-filler and planning aid rather than a substitute for professional safety judgment.

What the Source Says

The article says JESS, short for Journalist Expert Safety Support, produces advice on physical safety, digital security, legal risks, and field precautions once a user describes an assignment. CJR reports that the tool was conceived by Gina Chua and former Reuters bureau chief Mike Christie, and developed with the ACOS Alliance and the Journalism Protection Initiative. It says Kate Parkinson spent at least a hundred hours stress-testing the system and tailoring plans around identity-specific risks such as race, gender, age, and immigration status. The article also reports that JESS can hallucinate, that one tester found some earlier answers overwhelming or off-base, and that regular updating remains necessary. On privacy, the piece says chats are not used to train models, are not stored on servers by default, and are kept for only twenty-four hours in the user's browser unless the user chooses server-side history. It also says the system is designed to draw from a specific corpus and annotate answers with source links.