{
  "id": 6575908,
  "title": "An AI Assistant Handed Me 6 'Real' Invoice Samples. 5 Were Completely Made Up",
  "url": "https://urgent.news/2026/09/10/an-ai-assistant-handed-me-6-real-invoice-samples-5-were-completely",
  "topic": "ai",
  "section": "AI",
  "published": "2026-09-10T11:05:09.000Z",
  "source": {
    "name": "Dev.to",
    "slug": "dev-to",
    "url": "https://dev.to/donerightlabs/an-ai-assistant-handed-me-6-real-invoice-samples-5-were-completely-made-up-3ddo"
  },
  "original_language": "en",
  "account": "I was working on a parser for Vietnamese electronic invoices, aiming to convert the XML files into a standardized JSON format. To ensure accuracy, I asked an AI assistant to gather sample invoice XML files from various providers and a tax authority standard file. My parser rejected five of the six sample files due to discrepancies in tag names. Upon further investigation, I discovered that the data used in the sample files was simulated test data, not actual invoices. The AI assistant had created these samples based on generic tag names, not the actual government-mandated structure required for Vietnamese e-invoices. My parser correctly identified the five simulated files as fictional, while only accepting the one that used the real government-mandated tags. Later, when actual invoices started coming in, three bugs were discovered. Firstly, a privacy leak issue occurred when testing a utility bill invoice, as the buyer's name, address, and email were exposed in the raw fallback mode. Secondly, the parser failed to recognize the buyer's individual name tag, resulting in the processing of invoices without this information. Lastly, a bug emerged when dealing with an older invoice schema version, causing the invoice number and issue date to be silently returned as null. These issues were not detected by the AI-generated sample invoices, highlighting the importance of using real data for robust testing and error handling.",
  "summary": "I was building a parser for Vietnamese electronic invoices — the XML files tax authorities require for every VAT invoice issued in Vietnam. The goal: read a file from any provider, whoever issued it, and normalize it into one clean JSON shape. To sanity-check it against real-world variation, I asked another AI assistant to help me gather sample invoice XML files from a handful of different…",
  "key_points": [
    "AI assistant generated 6 invoice samples, 5 fabricated",
    "Parser rejected 5 sample files due to tag discrepancies",
    "Real invoice accepted, 3 bugs discovered in actual invoices"
  ],
  "editors_take": null,
  "illustration": null,
  "coverage": {
    "outlets": 1,
    "also_reported_by": []
  },
  "ai_generated": true,
  "disclaimer": "Summaries, key points and the editor’s take are written by software from other outlets’ reporting and may contain errors — always check the linked original."
}