Regulation — federal / defense
DoD Instruction 5000.02 — Adaptive Acquisition Framework + AI sustainment
DoDI 5000.02 governs how a program-of-record is acquired + sustained. The AI-sustainment implication: AI systems in the operations + sustainment phase carry the same lifecycle diligence obligations as the hardware they support. Silent drift in a sustainment-AI is a sustainment-phase failure — not a research-phase excuse.
What the instruction actually asks (paraphrased)
"Program managers will implement a lifecycle sustainment strategy that addresses reliability, availability, maintainability, and supportability of all system elements — including software-defined and AI-inclusive elements — from initial operational capability through disposal."
DoDI 5000.02 — representative sustainment-strategy expectation
"AI-inclusive systems will have documented performance-monitoring plans + drift-detection criteria + retraining triggers, incorporated into the program's product-support arrangement + reviewed at each acquisition milestone."
DoDI 5000.02 — representative AI-inclusive-sustainment expectation
What this means in plain English
Once an AI-inclusive system enters the sustainment phase, three obligations apply:
- Performance monitoring plan. Documented, in the program's product-support arrangement, reviewed at milestones. Not "the vendor watches it."
- Drift-detection criteria. Named thresholds + severity classification + response protocol. The A/B/C determination language directly maps here.
- Retraining triggers. When does the program-office force a model refresh? The 3 fix-first items in the Snapshot include this.
The instruction is not about the AI being perfect. It is about the sustainment discipline being real and defensible.
What triggers the exposure in the sample
13 drift events across the 90-day sustainment window, 12 of them high-severity, none surfaced by the vendor's fleet-readiness dashboard. Under DoDI 5000.02 lifecycle obligations, this pattern is precisely what the performance-monitoring plan is supposed to prevent — and precisely what the acquisition-milestone reviewer is trained to ask about.
What the $499 Snapshot shows against this instruction
- Documented drift-detection thresholds + severity classification — the "drift-detection criteria" evidence
- Dated first-drift-day + trajectory — the performance-monitoring plan operationalized
- 3 fix-first items including retraining-trigger recommendation — the retraining-trigger evidence
- Signed independent-verifier declaration — the record the acquisition-milestone reviewer will look for
See the reproducibility drill supporting sustainment-record production →
$499 Snapshot. 3 business days.
Sustainment-phase AI discipline evidence on your program's actual AI surface + 3 fix-first items + signed declaration.
Buy $499
Snapshot credit applies to Baseline ($2,500) or Enterprise Attestation ($35-55K) upgrade within 30 days.