{
  "id": 15723821,
  "title": "We need AI governance, not superintelligence hype",
  "url": "https://urgent.news/2026/10/20/we-need-ai-governance-not-superintelligence-hype",
  "topic": "ai",
  "section": "AI",
  "published": "2026-10-20T08:11:54.000Z",
  "source": {
    "name": "ITWeb",
    "slug": "itweb",
    "url": "https://www.itweb.co.za/article/we-need-ai-governance-not-superintelligence-hype/JN1gP7OAQWRqjL6m"
  },
  "original_language": "en",
  "account": "Many discussions surrounding artificial intelligence tend to focus on the concept of superintelligence, often overlooking critical questions about the nature of intelligence itself. These debates occasionally paint a picture of machines outperforming humanity, breaking free from our control, or dictating our future. However, these lofty concerns about superintelligence may divert attention from more pressing matters. For instance, who governs the infrastructure that powers AI? Who grants permission for its deployment? Who assesses its outputs? Who holds accountability when things go awry? The answers to these questions are pertinent and relevant to current developments in AI.\n\nSuperintelligence claims, in my view, rest on an incomplete understanding of what intelligence truly signifies. The term \"AI\" is a broad label that encapsulates various techniques and systems designed for prediction, classification, language generation, and action. Their abilities and constraints markedly differ. By treating all these different forms as a singular, all-encompassing entity, we fail to acknowledge these disparities. The term \"superintelligence\" risks reducing multiple capacities into a single attribute that can be arbitrarily enhanced.\n\nThe most significant existential threat may not stem from machines gaining superintelligence, but rather from leaders with unchecked ambition and inadequate judgment managing powerful technologies. Recent impressive performances of AI models on specific tasks do not provide a definitive answer to the philosophical question of what constitutes superintelligence.\n\nBehind these remarkable performances are complex networks of data centers, energy sources, training data, software integrations, human evaluations, and individuals tasked with correcting errors. The success of these systems is also contingent on reliable records, institutional knowledge, and staff expertise. Therefore, attributing these achievements to a self-sufficient machine can be misleading. The \"intelligence\" attributed to AI should be viewed as a product of the extensive infrastructure supporting it, rather than the machine alone.\n\nDescribing AI as unstoppable can lead to a dangerous race between superpowers and Big Tech firms. In such a scenario, questions about evidence, cost, and accountability might appear as resistance to progress. The Y2K scare is an example of how exaggerated predictions can make a repairable engineering problem seem beyond human control. Like the date-handling issues back then, AI safety narratives can obscure human agency. We need to ask: which failures are preventable? What corrective measures are necessary, who must implement them, and who should be held accountable if they are overlooked?\n\nAdditionally, warnings about future existential risks should not absolve Big Tech leaders from scrutiny of their current financial performance. These executives must still demonstrate the value of their investments. Furthermore, we should be wary when tech billionaires hold significant influence over political leaders. The industry's technical expertise is valuable, but it does not grant them the right to dictate the terms of public debate or policy.\n\nThe danger of regulatory capture is real - oversight bodies can inadvertently serve those they are supposed to scrutinize. The industry's technical prowess should not equate to the right to shape public discourse or policy. Historical lessons from tobacco and climate change sectors warn us of such pitfalls. Big Tech should not be allowed to set the terms of public debate or policy.\n\nIn 2024, the US Cyber Safety Review Board found an email breach preventable and criticized a major Big Tech provider's security culture. Although recent AI agent behavior incidents do not establish superintelligence, they do highlight the need for rigorous independent testing, incident reporting, and consequences for preventable failures. Boards must establish protocols for halting launches and allow staff to challenge unsafe decisions. Speculation about superintelligence should not detract from these immediate tests of leadership and oversight.\n\nUltimately, we do not need to definitively determine whether machines will become superintelligent to demand better judgment from those governing them today. The greatest existential threat may arise when powerful technologies are governed by leaders who lack restraint, negotiation, and reconsideration skills. The Cuban Missile Crisis serves as a powerful reminder of how decisions regarding power, conflict, and technology can lead societies to the brink of disaster. To avoid such a fate, we must insist on better judgment from those currently governing AI, even if we cannot predict whether machines will achieve superintelligence.",
  "summary": "AI safety cannot become a private understanding between the tech and political elites, with oversight serving those it should scrutinise.",
  "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."
}