{
  "id": 9758163,
  "title": "On Anthropic’s AI Misuse Report",
  "url": "https://urgent.news/2026/09/25/on-anthropics-ai-misuse-report",
  "topic": "ai",
  "section": "AI",
  "published": "2026-09-25T11:07:22.000Z",
  "source": {
    "name": "Schneier on Security",
    "slug": "schneier-on-security",
    "url": "https://www.schneier.com/blog/archives/2026/09/on-anthropics-ai-misuse-report.html"
  },
  "original_language": "en",
  "account": "Anthropic recently released a comprehensive report examining various misuse cases of its AI system, Claude. Over 100 findings were condensed into a digestible format by Daniel Meissler. Some key aspects highlighted in the report include the expanding capabilities of AI agents in reconnaissance, exploitation, data theft, propaganda production, surveillance workflows, and research tasks, with human operators overseeing target selection, goal setting, and reviewing critical outputs.\n\nThe report sheds light on attackers leveraging AI technology to streamline credential theft, cloud compromise, phishing attempts, vulnerability research, and extracting sensitive data from affiliated organizations. It also discusses the utilization of persistent agent memory in influence operations, which involves creating fake news sites, creating fabricated journalists, developing synthetic personas, conducting political profiling, and generating large-scale multilingual content. However, the report notes that an abundance of content often results in minimal genuine engagement.\n\nAnthropic's report also delves into surveillance and repression scenarios, encompassing automated dossier creation, biometric and communications analysis, transnational targeting, coercive recruitment tactics, and systems that persistently function locally even after model access is revoked. Additionally, the report reveals potential dual-use risks related to biological and weapons applications. While AI contributes to advanced scientific and military advancements, the report does not conclusively confirm the completion of biological weapons development or operational deployment in battlefield settings.",
  "summary": "Earlier this month, Anthropic published a long report detailing all of the Claude misuses it detected. Daniel Meissler usefully summarized the report into 117 findings. A few of the highlights: AI agents increasingly handled reconnaissance, exploitation, data theft, propaganda production, surveillance workflows, and research while humans selected targets, set goals, and reviewed important…",
  "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."
}