{
  "id": 1753545,
  "title": "My Data Pipeline Caught an AI Fabricating a Source. Here's What 19 Years of Safaricom's Financial Data Taught Me About Trust.",
  "url": "https://urgent.news/2026/08/18/my-data-pipeline-caught-an-ai-fabricating-a-source-heres-what-19",
  "topic": "ai",
  "section": "AI",
  "published": "2026-08-18T16:50:03.000Z",
  "source": {
    "name": "Dev.to",
    "slug": "dev-to",
    "url": "https://dev.to/derrickryangiggs/my-data-pipeline-caught-an-ai-fabricating-a-source-heres-what-19-years-of-safaricoms-financial-2ebe"
  },
  "original_language": "en",
  "account": "After constructing a data pipeline for Safaricom PLC, Kenya's largest telecom operator, the author encountered a significant issue - AI fabricating a source. The pipeline was designed to analyze 19 years of Safaricom's financial data, spanning from FY2008 to FY2026, but the AI-generated source created incorrect numbers with citations from Reuters and CNBC Africa. These citations were not the original sources, as neither outlet had ever reported such numbers. The author found this experience to be a valuable lesson in data engineering, highlighting the importance of trust in data pipelines. Safaricom PLC publishes some of the most detailed segment-level financial disclosures in East Africa, but these data are scattered across disconnected PDFs, making it difficult to analyze trends and compare data across years. The author's project, Safaricom Intelligence, transformed this data into a versioned, queryable BigQuery dataset that refreshes automatically every week, allowing users to easily query and visualize the data. The project used Terraform for infrastructure provisioning, Airflow for orchestration, pdfplumber for PDF extraction, dbt Cloud for data transformations, and Looker Studio for visualization. Despite the pipeline's success in providing accurate data, the author faced challenges when trusting the data, as fabricated sources could appear trustworthy due to accompanying citations. To address this issue, the author implemented a rule to verify data against primary sources in order of reliability, starting with press releases and moving down to annual reports. This rule helped catch erroneous numbers and ensure the data's trustworthiness. Ultimately, the pipeline provided valuable insights into Safaricom's M-PESA business growth, which now makes up a significant portion of the company's revenue.",
  "summary": "Partway through building a data pipeline on Safaricom's public financial disclosures, one of my own extraction passes handed me two numbers for Ethiopia's EBIT loss, and a citation to back them up. The citation credited the quote to Reuters and CNBC Africa. Neither outlet had ever said it. The numbers were wrong, and something along the way had generated a source to make the wrong numbers look…",
  "key_points": [],
  "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."
}