Incident handoff4 min readField note 11

How to hand off an AI data exposure investigation to a reviewer

Build a case another analyst can reproduce. Use a Claude paste and a ChatGPT spreadsheet receipt to practice evidence selection, assignment and export verification.

Selected Claude paste and ChatGPT file receipts in the Quarter-end AI evidence review case.
The idea in focus

Two observations share one fictional Northstar case. Each keeps its own destination, action and source.

Field note / 11
In brief

Give the next reviewer the receipts and the questions that remain open.

  • Preserve each observation’s source and time.
  • Assign the next decision to a named owner.
  • Check the copy of any export you hand over.
In this article

Frame the question before opening a case#

A Finance manager asks whether a quarter-end workbook or a credential reached an AI service. The outgoing analyst has two observations on the same device. Before creating a case, check the shared queue for an existing owner. Write the decision the reviewer needs to make. In this example it is whether either observation warrants incident escalation, and what evidence would support that decision.

Shared case queue containing Finance report exposure and sensitive paste reviews.
In the fictional Northstar queue, inspect existing cases before opening another review.

Select the receipts that address the question#

Create a case in the demo and select the Claude paste and the spreadsheet receipt. Open the timeline and check the account, hostname, time and destination on each row. The events can belong in one review because they answer the same intake question. Their proximity on a timeline does not establish that the pasted text came from the workbook.

The new case form has two evidence items selected, one paste and one file-related observation.
Select the two relevant observations individually so the reviewer can reconstruct the scope.

Write down what each receipt establishes#

Open the Claude observation. Capture its detector category and the gesture recorded on the supported page. Then read the spreadsheet observation separately. A concise handoff note could read “Claude paste contained a key-like pattern. ChatGPT recorded a related response for Q4_client_report.xlsx. Provider acceptance and credential validity remain open.” This gives the next analyst two testable questions without copying a secret into the case description.

Detail view of a selected observation with its source, destination and finding information.
Read the source observation before writing the handoff note.

Set the assignee and next decision#

Assign the case to the person responsible for the next decision. Include the originating timezone in any separate incident note, especially if teams hand over across shifts. Ask the credential owner to review validity through the normal incident channel. Ask the data owner which classification applied to this workbook version. Record who will bring each answer back and when.

Case edit form with the reviewer selected and the case still open.
Save the assignment and leave the case open while questions remain.

Check that the handoff reaches the reviewer#

Switch to the reviewer persona in the public demo and select the assigned-to-me view. Reopen the case from that queue and check both receipts. For a real shift change, ask the receiving analyst to acknowledge the specific open question in your incident system. Access to the case and ownership of its next action are separate checks.

Assigned-to-me case queue with the reviewer persona selected.
The receiving reviewer can find the case in their assigned queue.

Verify the export that accompanies the handoff#

In the demo evidence workbench, select the same receipts, export the sample case and choose the downloaded file in the verifier. A match establishes that the bytes checked match the recorded digest. It says nothing about what the AI provider retained. In an operational handoff, preserve the reference digest independently, identify who exported the records and log custody of the copy. The public exercise uses fictional records and a browser-local workflow.

Export digest and successful verification result for the selected sample case file.
The demo checks the selected export against its recorded digest in the same browser session.
Read the primary sources

Sources & further reading

  1. Try the shared case queue
  2. Investigation guide
  3. What proves an AI document upload?
← Back to all articlesBack to top ↑
A CLEAR PLACE TO START

Plan your AI visibility pilot.

Create a brief for your team. Your answers stay in this browser and are included in your download.

AI tools in scope
More supported tools

This creates a planning document. It does not enroll devices or submit a request.