# A US federal court fined two attorneys and their firm $5,000 for a brief citing six non-existent cases generated by ChatGPT, noting that using a reliable AI tool is not itself improper

Page: https://flyvolo.ai/en/changes/ev-20230622-paralegal-1
Date: 2023-06-22 · Stage: Constraint
Can this record move a task judgement? yes
Occupation: https://flyvolo.ai/en/careers/paralegal

## Scope

US federal court (S.D.N.Y., Judge Castel). The sanction rested on failing to verify and on standing by the citations once challenged. Widely cited in later professional-conduct guidance.

## Source

- Mata v. Avianca, Inc. (S.D.N.Y., 22 June 2023) — opinion and order on sanctions, via CourtListener — https://www.courtlistener.com/docket/63107798/54/mata-v-avianca-inc/ (primary source)

## What this means

A constraint record that creates the linked task: a federal court fined lawyers $5,000 not for using ChatGPT but for filing six invented cases without checking, and standing by them once challenged. On the deployment layer it fixes where the tools may be used in litigation — only with someone confirming every citation exists — and that confirming step is the 'checking what the tools produced' work now landing on paralegals.

## What it does not show yet

One sanction order in one US district says nothing about how firms pay for verification: whether it becomes a billable paralegal skill or unpaid diligence is a firm-by-firm decision the opinion does not touch. It does not show fabricated citations are common, only that the cost of one slipping through is now on record — and it applies to court filings, not to contract or research work.

## How to verify it yourself

Ask your supervising attorney whether the firm has a written rule on who verifies citations in machine-assisted drafts and whether that time is recorded to a matter code; no rule means the task exists but has not been assigned, and non-billable time means it has been assigned to you for free.


More: https://flyvolo.ai/llms.txt
