A response winds down the way most do. The final report is filed, the office is handed back, the team disperses to the next emergency. But the records remain. Names, locations, biometric scans, the precise coordinates of who was most vulnerable and why. We plan the entry into a crisis in obsessive detail. We rarely plan the exit for the data we created, and so it lingers on a server nobody owns, in an account nobody monitors, protected by a duty of care that quietly expired with the contract.
This is a governance gap, not a technical one. We treat data collection as the deliverable and data retirement as an afterthought, if we think of it at all. The risk does not end when the project does. A list of displaced families is an asset to whoever fled, and a target to whoever they fled from. The longer it sits past its purpose, the more it shifts from protection to liability. Good intentions at the point of collection do nothing to govern a dataset three years after the people who collected it have gone.
The better version writes the ending at the beginning. Every dataset gets an owner of record who survives the project, a retention period tied to the actual purpose, and a deletion date that someone is accountable for honoring. We decide before we collect what gets archived, what gets handed to a successor under what terms, and what gets securely destroyed the moment it is no longer needed. Minimisation helps here too, because the data we never gathered is the data we never have to guard. A shorter list is a smaller risk to carry into the years after the cameras leave.
So the next move fits inside the proposal we are already writing. Add a data lifecycle plan beside the budget: who holds it, for how long, who deletes it, and how we prove they did. Treat the closeout of a dataset as seriously as the closeout of a bank account, with a sign off that says this is gone and here is the evidence. The people who trusted us with their information did not consent to being remembered forever by a system that stopped caring for them. Protection is a promise that has to outlast the funding, or it was never protection at all.