Rule — legal-specific
FRCP 37(e) — spoliation duty on AI-driven case-intake decisions
Federal Rule of Civil Procedure 37(e) governs failure to preserve electronically stored information. When AI drives intake, routing, or case-value estimation, the model version + inputs + outputs at decision-time are ESI. Once litigation is reasonably anticipated, they must be preserved. Failure to preserve carries sanctions up to and including an adverse-inference instruction.
What the rule actually says
"If electronically stored information that should have been preserved in the anticipation or conduct of litigation is lost because a party failed to take reasonable steps to preserve it, and it cannot be restored or replaced through additional discovery, the court: (1) upon finding prejudice to another party from loss of the information, may order measures no greater than necessary to cure the prejudice; or (2) only upon finding that the party acted with the intent to deprive another party of the information's use in the litigation may: (A) presume that the lost information was unfavorable to the party; (B) instruct the jury that it may or must presume the information was unfavorable to the party; or (C) dismiss the action or enter a default judgment."
Fed. R. Civ. P. 37(e) — the operative text
What this means in plain English
Three obligations any PI firm using AI in case intake now carries:
- Identify AI-driven decisions as ESI. The triage score, the lane routed, the model version at decision-time, and the input snapshot are all ESI. A screenshot of a dashboard is not a preservation-grade record.
- Preserve from the moment litigation is reasonably anticipated. For a plaintiff PI firm, "reasonable anticipation" starts as early as the intake itself — the firm is holding the case for the client, not defending it.
- Reproduce the decision on demand. If defense counsel, opposing expert, or the judge asks for the AI-generated triage record for a specific matter, the firm needs a defensible reproduction procedure — not a "we think it was routed to lane X" narrative.
Intent to deprive is the trigger for the harshest sanctions. But the "reasonable steps" prong applies regardless of intent, and the modal AI vendor stack does not retain decision records in a preservation-grade form.
What triggers the exposure in the sample
C3 SIU/fraud-review lane rate rose from 18% to 71% over 45+ silent days. Roughly 430 potentially-misrouted C3 intakes sat in the wrong lane during the drift window. When any one of those clients — or a class-adjacent group — later challenges the routing, the firm needs to produce the model version, the input snapshot, and the decision record for that specific intake. If those records were not preserved, FRCP 37(e) applies.
The rule does not care whether the firm meant to lose the records. It cares whether reasonable steps were taken.
What the $499 Snapshot shows against this rule
- Per-segment distributional-shape analysis — surfaces the drift window that defines preservation scope
- Model-version pin + decision-hash lookup demonstrated on 5 sample intakes — the "reasonable steps" audit trail
- 3 fix-first items including litigation-hold + retention-record fix — the preservation starting point
- Independent-verifier signature — stance that survives defense counsel's "was your monitoring independent" question
See the 5-decision reproducibility drill →
$499 Snapshot. 3 business days.
Independent-verifier determination scoped to your firm's 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.