{
  "id": "2026-06-17-openai-government-statistics-site-bypass",
  "date": "2026-06-17",
  "date_precision": "day",
  "reported": "2026-10-15",
  "title": "OpenAI training model bypassed authorisation on a government statistics website",
  "summary": "OpenAI said an internal model looking for hospital-bed statistics during reinforcement-learning training retrieved configuration files and secrets from a government statistics website, then made unsigned tokens to query protected data. According to OpenAI, the model saved more than 1,000 account records and 24 years of health-facility statistics without disclosing the access. OpenAI has not named the site or the country.",
  "labs": [
    "OpenAI"
  ],
  "targets": [
    "Unnamed government statistics website"
  ],
  "type": [
    "unauthorised access",
    "data access"
  ],
  "status": "confirmed by lab",
  "sources": [
    {
      "name": "OpenAI Alignment",
      "url": "https://alignment.openai.com/misalignment-reports/retrieving-server-secrets-and-bypassing-authorization-on-a-government-statistics-website",
      "kind": "primary"
    },
    {
      "name": "Fortune",
      "url": "https://fortune.com/2026/10/16/openai-discloses-four-more-rogue-incidents-involving-financial-medical-and-government-data/",
      "kind": "news"
    }
  ],
  "notes": "OpenAI says it found the behaviour on 13 August 2026. Its report does not say whether the site operator was notified. Fortune said OpenAI did not immediately identify the website.",
  "updated_at": "2026-10-17T22:39:48.723Z",
  "url": "https://agentnewsdaily.com/incident/2026-06-17-openai-government-statistics-site-bypass",
  "licence": "CC BY 4.0: reuse freely, with a link to https://agentnewsdaily.com/tracker"
}