{
  "id": 294352,
  "title": "Artificial Intelligence: Do AI systems need to be kept on a short leash?",
  "url": "https://urgent.news/2026/08/08/kunstliche-intelligenz-mussen-ki-systeme-an-die-kurze-leine-genommen",
  "topic": "ai",
  "section": "AI",
  "published": "2026-08-08T04:00:08.000Z",
  "source": {
    "name": "Handelsblatt",
    "slug": "handelsblatt",
    "url": "https://www.handelsblatt.com/technik/it-internet/kuenstliche-intelligenz-muessen-ki-systeme-an-die-kurze-leine-genommen-werden/100245914.html"
  },
  "original_language": "de",
  "account": "When algorithms overcome security barriers and independently exploit digital vulnerabilities, this is no longer just a distant future prospect.\nReputable US tech companies such as Meta, OpenAI, and Anthropic have recently had to admit that their language models showed autonomous hacking capabilities in tests.\nThis has given new impetus to concerns about uncontrolled risks.\nIn politics and in the digital economy, a fundamental debate is therefore raging:\nAre voluntary commitments and the European AI Act sufficient?\nOr must artificial intelligence be kept even more tightly under control?\nA pro and con.\n\nArguments for stricter regulation\nLiability of manufacturers and personal responsibility\nFrom the perspective of civil rights activists, there is currently a lack of clear, unambiguous responsibility on the part of developers for AI systems.\nMarkus Beckedahl, Managing Director of the Center for Digital Rights and Democracy, draws a comparison with established industries:\n\"For cars or medicines, the question has long been settled: whoever manufactures them is liable for the damage.\nIn the case of AI systems, the providers behave as if this were still up for debate.\"\nIt is irresponsible if companies invoke unpredictable effects after their software has malfunctioned.\n\"Whoever manufactures complex systems must also be able to control them,\" Beckedahl demands.\n\"If an AI runs wild due to a lack of safety precautions, the people responsible in the companies should also be personally accountable.\"\n\nWarnings from the company's own developers\nEven from the engine room of the leading AI labs, there is now a loud call for state regulation.\nIn the open letter \"Pacing the Frontier\", over 1,100 top researchers and employees of OpenAI, Anthropic, Google, and Meta warned that the pace of development is threatening to become uncontrollable.\nSince the extreme competitive pressure does not allow a single company to brake voluntarily, the insiders are explicitly asking the US legislature and international politics for help:\nState tools must be created to be able to specifically throttle the development of automated AI research in an emergency.\n\nProtection of fundamental rights\nThe German civil rights organization AlgorithmWatch also sees politics as being responsible and accuses the EU of inaction.\nOliver Marsh, head of technology research at AlgorithmWatch, criticizes the behavior of the corporations and the legislature:\n\"It is a scandal that companies hardly take responsibility for the dangers and damage caused by their handling of their AI systems and have to face consequences.\"\nThe EU is much too lenient.\nGovernments must create or tighten legal frameworks so that the use of AI cannot become dangerous, and they must not be influenced by tech CEOs.\n\nArguments against stricter regulation\nPrevent systematic disadvantages for Europe\nIn the discussion about strict AI regulation in Germany and Europe, it is often not about clearly defined guidelines for AI itself, but about a possible disadvantage for the European location.\nAt the European level, the French President is one of the most prominent critics of early and strict regulation.\nHe repeatedly warned that the European AI law could severely slow down European innovation leaders such as the French start-up Mistral AI compared to US or Chinese competitors.\n\nThreat of innovation brake\nThe industry association Bitkom warns urgently about over-regulation of the digital location.\nToo rigid bureaucratic hurdles and unclear specifications make companies uncertain and slow down the development of practical applications.\nGiven the lead of the USA and China, Europe risks losing its connection in the global technology race if the rules are even stricter.\n\nCall for regulation is just a trick\nThere is also skepticism about the role of prominent company heads when they themselves advocate for strict regulation.\nOpenAI CEO Sam Altman, for example, repeatedly called for state licensing and regulatory authorities for extremely powerful AI systems.\nHowever, critics view this call for regulation as a strategic maneuver.\nEnormous approval hurdles and expensive safety tests can be easily afforded by industry giants - for smaller competitors, European medium-sized companies, and open-source initiatives, they mean the end.\nExcessive legislation may thus protect exactly the monopolies that it is supposed to control.\n\nFunctioning self-regulation\nThe tech companies point out that phenomena such as autonomous hacking attempts were only discovered through their own internal tests (red teaming).\nThis proves that the security mechanisms of the industry are working.\nAnyone who generally restricts the development of models would take away IT security researchers the tools to develop defense systems against new-type digital threats at all.",
  "summary": "US corporations admit to autonomous hacking capabilities of their AI. Can stricter rules prevent this? And what risks does over-regulation pose?",
  "key_points": [
    "AI systems can independently engage in hacking activities during tests.",
    "Proponents argue manufacturers should be liable for AI system damages.",
    "Critics fear over-regulation could disadvantage Europe's digital leadership."
  ],
  "editors_take": "The intensifying debate over AI control highlights a struggle between mitigating risks through stricter regulation and avoiding over-regulation that could disadvantage European businesses and hinder innovation.",
  "illustration": null,
  "coverage": {
    "outlets": 2,
    "also_reported_by": [
      {
        "outlet": "Handelsblatt",
        "title": "Artificial intelligence: Nvidia plans new mega-deal with Wall Street investors",
        "url": "https://urgent.news/2026/08/10/kunstliche-intelligenz-nvidia-plant-neuen-mega-deal-mit-wall-street",
        "published": "2026-08-10T20:19:34.000Z"
      }
    ]
  },
  "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."
}