{
  "id": 12575506,
  "title": "“계정 10센트, 딥페이크 80달러”…다크웹서 헐값에 팔리는 AI 범죄 도구",
  "url": "https://urgent.news/2026/10/07/10-80-ai",
  "topic": "tech",
  "section": "Tech",
  "published": "2026-10-07T05:46:29.000Z",
  "source": {
    "name": "Hankyoreh",
    "slug": "hankyoreh",
    "url": "https://www.hani.co.kr/arti/international/america/1281341.html"
  },
  "original_language": "ko",
  "account": "A recent surge in dark web markets selling AI criminal tools has raised concerns among cybersecurity experts. The Hidden Internet Research Center released a report titled 'Underground AI Services' on the 6th, revealing that the artificial intelligence cybercrime economy has already arrived, with a massive commercial ecosystem forming in the past six months. In the past six months, the market saw a 28-fold increase in postings related to AI cybercrime tools and services, from less than 50 per month at the end of last year to over 1,400 in February this year.\n\nThe analysis of 4,000 posts across 20 dark web channels, 7 Telegram channels, and 5 markets found that the most common offerings were weaponized large language models (LLMs), identity theft, malicious code, and phishing attacks. Weaponized LLMs, which accounted for 24% of the 4,000 posts (888 listings), were primarily generated by modifying open-source AI models to enable the creation of phishing emails and malware. One such service, named 'WormGPT', started in 2023 but was forced to shut down after two months due to issues. However, other developers have since launched similar services, with the latest version reaching 5.2.\n\nThe most lucrative AI cybercrime services were identity theft (15 dollars for a fake driver's license), creating AI-generated deepfake videos (80 dollars), and providing AI-based identity theft services (up to 6,350 dollars for a package combining fake IDs, utility bills, bank statements, and real-time deepfake videos to bypass financial institution's KYC processes). AI-based phishing call centers offering calls to 120 people in 25 languages and AI-based hacking tools and attack infrastructure services were also popular, accounting for 18.95% and 25.6% of the posts, respectively. Open-source models less regulated than others like OpenAI and Anthropic have been found to be more frequently exploited for criminal purposes.",
  "summary": null,
  "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."
}