Legacy System Decommissioning: Writing the Disposition Plan for a Government System After Migration

Many modernization plans stop at go-live and leave the old system running. What GAO, NIST, and the National Archives expect when a legacy system is retired, and a shutdown sequence with exit conditions a program office can use.

When a replacement system goes live, the legacy system is usually still running. It still has user and service accounts, network rules that let other systems reach it, certificates and keys that other systems trust, and storage that holds decades of records. Until it is switched off, the program pays to operate two systems and has to defend both.

This post covers the last phase of a legacy migration: the disposition plan, which describes how the old system is taken out of service. It summarizes what federal auditors and standards expect, and gives a shutdown sequence with an exit condition for each step.

Most modernization plans stop before the shutdown

In July 2025, the Government Accountability Office (GAO) reported in GAO-25-107795 on the 11 federal legacy systems most in need of modernization. According to the agencies, those systems were about 23 to 60 years old and together cost approximately $754 million a year to operate and maintain.

GAO checked each modernization plan for three elements: milestones, a description of the work, and details on the disposition of the legacy system. Agencies had plans for nine of the 11 systems, and three of those plans contained all three elements. One of the complete plans, at the Department of the Interior, included a decommissioning plan with pre-decommissioning and post-decommissioning requirements and detailed decommissioning activities.

The same report cites an August 2024 finding by the Treasury Inspector General for Tax Administration: 105 of 107 Internal Revenue Service legacy systems that had been identified for retirement did not have specific decommissioning plans.

A retirement date with no plan behind it tends to slip. GAO describes a Department of Defense system that was announced in March 2000 for retirement in October 2002. The deadline was missed, the system was not replaced, and it was still on GAO’s list in 2025.

What federal guidance asks for at disposal

Three sources cover most of what a disposition plan has to address. Each one produces a specific document that an assessor or auditor can ask to see.

SourceWhat it asks forEvidence it produces
NIST Special Publication 800-37, Revision 2, Task M-7“Implement a system disposal strategy and execute required actions when a system is removed from operation.” The system owner is responsible.Disposal strategy, updated system component inventory, updated security and privacy plans
NIST Special Publication 800-88, Revision 2 (September 2025)An agency media sanitization program, with a validation decision on whether each sanitization was effectiveA certificate of sanitization for each storage item
36 CFR 1236.14 (National Archives regulation)If records must be kept longer than the life of the system, plan their migration to a new system before the current one is retired, and keep each record linked to its metadataRecords migration plan and the records officer’s approval

The discussion under Task M-7 lists further actions that are easy to overlook: inventory and tracking systems are updated to show the removal, users and the owners of hosted applications are notified, and control inheritance relationships are reviewed. Other systems may inherit a control, such as audit log collection, from the one being retired.

Settle the records question first

Storage cannot be sanitized until someone with authority has confirmed that every record on it has either been migrated, transferred, or reached the end of its approved retention period. That person is the agency records officer, and the confirmation should be a signed entry in the plan.

This step is often missed because the migration team and the records staff are in different offices. A National Archives inspection of one defense agency found that program managers rarely informed the records officer when systems were decommissioned, and stated that this “significantly increases the risk of unauthorized disposition or other mismanagement of records.”

Data that was deliberately left out of the migration needs particular attention. If closed cases older than a cutoff date were not converted, they still have a retention period, and they need an archive with a tested way to retrieve them. Our post on cutover and reconciliation covers the conversion itself.

Show that nothing still depends on the old system

A legacy system can be retired only when no person, job, or external system still calls it. The interface inventory built before migration is the starting list (see our post on dependency analysis before a legacy migration). It should be checked against what happens on the network after go-live.

Put the system in read-only mode, log every inbound connection, and trace each one to an owner. Keep observing through at least one occurrence of every scheduled cycle the system supported, including quarter-end and fiscal year-end. A job that runs once a year will not appear in a 30-day observation period.

Remove trust, then sanitize storage

Before the hardware is touched, remove everything that lets the old system act or be trusted:

  • Disable user accounts and service accounts, including accounts the system holds on other systems.
  • Revoke certificates issued to the system and remove its keys from the trust stores of its partners.
  • Rotate any shared secret or key that the old system and a surviving system both hold.
  • Delete firewall rules, name service entries, and scheduled jobs that reference it.

Sanitization follows. Revision 2 of SP 800-88 moved its emphasis from device-level instructions to an organizational program, and it treats cryptographic erase as an acceptable technique when its conditions are met. Cryptographic erase means sanitizing the keys that encrypt the stored data, so that the data can no longer be decrypted. For federal agencies, the encryption module must be validated to the current FIPS 140 standard.

The guideline adds a caution that applies directly to old systems. It says cryptographic erase “should not be trusted” on media “that have been backed up or escrowed unless the organization has a high level of confidence regarding how and where the keys were stored and managed” outside the media. A system that has run for 30 years often has key copies and backup tapes that nobody on the current team created. Backups, replicas, test environments loaded with production data, and copies held by contractors all belong in the sanitization scope.

An illustrative shutdown sequence

Consider a benefits case management system on a midrange server, replaced by a new application. All numbers below are illustrative.

StepExit conditionEvidence kept
1. Read-only fallback (90 days)Go-live defects closed, fallback no longer neededDecision record signed by the system owner
2. Dependency observation (through fiscal year-end)Zero unexplained inbound connections for 60 daysConnection logs with an owner for each caller
3. Records dispositionRecords officer approves the archive of 1.8 million unconverted closed casesSigned approval, retrieval test result
4. Trust removal23 accounts disabled, 14 certificates revoked, 9 firewall rules deletedChange tickets, revocation list entries
5. Sanitization42 drives and 310 backup tapes sanitized and validatedOne certificate of sanitization per item
6. CloseoutInventory updated, authorization terminated, support contract endedUpdated inventory and security plan

In step 2, the logs show three callers that were not in the interface inventory: a quarterly reporting job in a regional office, a monitoring probe, and an annual actuarial extract that ran once during the observation period. Each is redirected or retired before step 3 begins. If the system had been powered off after 30 days, the annual extract would have failed eight months later with nobody left who knew why.

How QuantumWorks approaches this

PRISM analyzes a legacy codebase and maps entry points, behavior, and dependencies, each anchored to file and line, which gives a team the first version of the list of things that call the old system. Sentinel helps organizations understand and change cryptography across many systems at once, which includes finding where a retiring system’s certificates and keys are still in use. CAM provides asset-level cryptographic evidence for assets in production after migration. Sentinel and CAM each have a provisional patent issued.

If your program has a go-live date and no written plan for switching off the system it replaces, we would be glad to talk. Contact QuantumWorks.

  • Legacy System Decommissioning
  • Disposition Plan
  • System Retirement
  • Media Sanitization
  • Records Management
  • NIST SP 800-88
  • Legacy Systems

Working on a hard technical problem?

We work with government organizations, research institutions, and technology teams across cybersecurity, software, and autonomous systems.