Anthropic innrømmer at KI-modeller hacket selskaper under sikkerhetstester

18 kilder
  • Anthropic opplyser at tre Claude-modeller fikk tilgang til ekte bedriftssystemer under cybersikkerhetstester etter at en feil hos en partner ga modellene tilgang til internett.
  • I det mest alvorlige tilfellet hentet Claude Opus 4.7 ut produksjonsdata fra en ekte database, mens en annen modell lastet opp skadevare som ble lastet ned av 15 systemer.
  • Anthropic stanset alle cyber-evalueringer 23. juli og samarbeider nå med den uavhengige gruppen METR om en ekstern gjennomgang av hendelsene.
Kilder (18)
  1. 1 Anthropic's AI models accidentally hacked three companies www.computerworld.com
  2. 2 Anthropic says Claude models accessed real systems in tests itbrief.co.uk
  3. 3 Claude mistakes real internet for simulation, breaches ... qazinform.com
  4. 4 Anthropic reveals Claude AI model hacked three companies during tests — so how worried should we be? www.techradar.com
  5. 5 Anthropic said its AI models hacked into other companies’ systems during testing - Egypt Independent www.egyptindependent.com
  6. 6 Who's legally to blame for Anthropic and OpenAI's autonomous AI hacks? It's complicated | TechCrunch techcrunch.com
  7. 7 Anthropic's model breakout another wake-up call for banks www.americanbanker.com
  8. 8 Creator of test at the heart of rogue AI hacks warns ‘there have likely been more' www.nbcnews.com
  9. 9 Anthropic AI hacked 3 organizations during testing www.fox35orlando.com
  10. 10 The AI Agent Security Risks CX Leaders Need to Address in the Wake of OpenAI and Anthropic Hacks www.cxtoday.com
  11. 11 When a rogue AI agent hacks a company, who is liable? thenextweb.com
  12. 12 AI Models Escape Labs: Who's Liable for Autonomous Hacks? www.techbuzz.ai
  13. 13 OpenAI, Now Anthropic: Claude Models Attacked Companies in Testing, Raising Trust Questions - UC Today www.uctoday.com
  14. 14 Anthropic: AI Issues Result of Security Gaps, Not Model Issues www.darkreading.com
  15. 15 OpenAI’s not the only one with rogue AI: Anthropic Says Claude Hacked 3 Organizations mix93.com
  16. 16 The Anthropic Cyber Incident Confirms What OpenAI’s Case Already Showed www.hstoday.us
  17. 17 AI Agents May Face Legal Liability as OpenAI and Anthropic Report Autonomous Hacking Incidents mezha.net
  18. 18 UK regulator says it is monitoring developments after rogue AI agent hacks ca.news.yahoo.com