“Be careful with confidential data” is good advice, but it does not help a writer at 6 p.m. with a deadline and a document open. They need a faster test.<\/p>\n
A traffic-light system gives one. Sort your content into green, amber, and red, and let the color decide which tools are acceptable.<\/p>\n
This is a starting framework. Your own compliance and legal teams set the final rules.<\/p>\n
Green content is already public or contains nothing sensitive.<\/p>\n
Examples:<\/strong><\/mark><\/p>\n Rule:<\/strong><\/mark> Any tool your organization has approved is fine. Even here, follow your company’s AI policy and disclosure rules.<\/p>\n Amber content is not public and would cause harm if exposed, but it does not identify individuals.<\/p>\n Examples:<\/strong><\/mark><\/p>\n Rule:<\/strong> <\/mark>Use only approved tools with clear written terms: no storage beyond processing, no training use, and an agreement your compliance team accepts.<\/p>\n Red content contains information about individuals, or is covered by a legal or contractual restriction.<\/p>\n Examples:<\/strong><\/mark><\/p>\n Rule:<\/strong><\/mark> Do not use any tool unless your privacy or compliance team has approved it for this content, with the right agreement in place, such as a BAA or DPA. For the most sensitive material, consider an option that never leaves the device. Where possible, remove identifiers first.<\/p>\n Writers often assume data without names is safe. It is not.<\/p>\n In the Weill Cornell Medicine incident disclosed this month, the data included medical record numbers and clinical details but not names, and it still became a reportable event. Medical record numbers are among the 18 identifier types in HIPAA’s Safe Harbor method.<\/p>\n Free text is the other trap. A single exact date or a rare condition in a narrative can point to one person.<\/p>\n If the answer to question 2 or 4 is yes, treat it as red. If question 3 is yes, treat it as amber at minimum.<\/p>\n Your classification may differ, and a draft can change color as detail is added.<\/p>\n\n
Amber: Confidential, But Not Personal<\/h2>\n
\n
Red: Personal or Regulated Data<\/h2>\n
\n
Why Red Is Wider Than It Looks<\/h2>\n
Five Questions to Sort Any Document<\/h2>\n
\n
A Quick Reference<\/h2>\n
\n\n
\n Document<\/th>\n Likely color<\/th>\n<\/tr>\n \n Published abstract or paper<\/td>\n Green<\/td>\n<\/tr>\n \n Draft manuscript, no patient detail<\/td>\n Amber<\/td>\n<\/tr>\n \n Clinical study report section with aggregate results<\/td>\n Amber<\/td>\n<\/tr>\n \n Case narrative with dates and locations<\/td>\n Red<\/td>\n<\/tr>\n \n Participant-level listing<\/td>\n Red<\/td>\n<\/tr>\n \n Regulatory response letter before submission<\/td>\n Amber<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n Where Trinka’s Confidential Data Plan Fits<\/h2>\n