{
  "id": 8900533,
  "title": "Microsoft Couldn't Answer That Question Under Oath",
  "url": "https://urgent.news/2026/09/21/microsoft-couldnt-answer-that-question-under-oath",
  "topic": "tech",
  "section": "Tech",
  "published": "2026-09-21T03:44:56.000Z",
  "source": {
    "name": "HackerNoon",
    "slug": "hackernoon",
    "url": "https://hackernoon.com/microsoft-couldnt-answer-that-question-under-oath?source=rss"
  },
  "original_language": "en",
  "account": "On June 9, 2025, Anton Carniaux, Microsoft's legal director for France, appeared before a French Senate commission investigating public procurement and testified under oath. When asked if he could guarantee French citizens' data held under French public contracts would never be handed to U.S. authorities without French consent, Carniaux admitted he couldn't. While this answer was honest and the second half didn't address whether such a demand had occurred, it highlighted a fundamental sovereignty issue. The room had questioned whether data could be handed over, not if it had already happened. This exchange encapsulates the core problem of sovereignty in one sentence: a company operating under U.S. jurisdiction cannot guarantee otherwise, regardless of server location or contract terms. The debate has since expanded from cloud storage to AI and intelligence. A new report shows 184 government-backed sovereign AI initiatives across 67 countries, with 41 launched in the first half of 2026 alone. McKinsey predicts the sovereign AI market could reach $600 billion by 2030, with 71% of surveyed executives and officials considering it an existential concern or strategic imperative. However, the urgency behind this trend stems from a distribution problem: only 32 countries host AI data centers, while over 150 do not. This leaves most of the world reliant on other jurisdictions for their intelligence. Sovereignty in AI deployments consists of four key aspects: infrastructure, data, model, and operational sovereignty. While infrastructure sovereignty is often purchased, it does not address the remaining three pillars. Building the entire stack yourself is not a complete solution, as it still leaves a single point of failure. Aphanarc offers a solution by atomizing compute, model execution, and data handling, preventing any single location from holding the complete picture. This approach addresses the critical need for structured dependency, ensuring no single party can compromise the system. Carniaux couldn't promise a yes or no answer, but Aphanarc's architecture can, making it a crucial consideration in AI procurement conversations.",
  "summary": "What was a procurement argument about email and storage is now a national-strategy argument about intelligence itself.",
  "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."
}