Anthropic: Moonshot AI nutzte Claude heimlich für Modelltraining

38 Quellen
  • Anthropic gibt an, dass Moonshot AI heimlich fast 300 000 Nutzeranfragen über 5 380 gefälschte Konten in Singapur und Japan an das Modell Claude Opus weitergeleitet hat.
  • Der am Donnerstag veröffentlichte Bericht besagt, dass Moonshot die Antworten zur Schulung seines eigenen Kimi-Chatbots in einem Prozess namens Destillation nutzte; auch DeepSeek und Xiaomi wurden genannt.
  • Die Ergebnisse folgen auf eine US-Regierungswarnung in dieser Woche, die sechs chinesische KI-Firmen beschuldigt, im industriellen Maßstab "Milliarden von Tokens" aus amerikanischen KI-Systemen extrahiert zu haben.
Quellen (38)
  1. 1 Moonshot Secretly Routed User Requests Through Claude, Anthropic Says www.bloomberg.com
  2. 2 Chinese AI Giants Accused of Sending Millions of User Queries to U.S. Models www.wsj.com
  3. 3 US accuses Chinese AI firms of 'malicious' copying of AI ... www.reuters.com
  4. 4 Anthropic exposes how Chinese AI firms try to steal LLM tech mashable.com
  5. 5 Anthropic accuses DeepSeek, Moonshot and MiniMax ... www.cnbc.com
  6. 6 China-Based Artificial Intelligence Companies Conducting ... www.cisa.gov
  7. 7 Moonshot Secretly Routed User Requests Through Claude, Anthropic Says - Bloomberg News www.tradingview.com
  8. 8 Anthropic accuses Moonshot AI of misusing Claude models: Bloomberg cryptobriefing.com
  9. 9 Detecting and preventing distillation attacks www.anthropic.com
  10. 10 What "Distillation" Actually Does for Chinese AI Companies yage.ai
  11. 11 Anthropic Exposes Industrial-Scale Distillation Attacks by ... rits.shanghai.nyu.edu
  12. 12 Competing AI strategies for the US and China www.brookings.edu
  13. 13 Anthropic Distillation Attacks: DeepSeek, Moonshot, MiniMax www.digitalapplied.com
  14. 14 US Lead in the AI Race With China Is Rapidly Narrowing www.bloomberg.com
  15. 15 The Attack That Looked Like Nothing at All: Anthropic's ... treblle.com
  16. 16 "We've identified industrial-scale distillation attacks on our ... www.reddit.com
  17. 17 China researchers using US AI models for defense systems. thecyberwire.com
  18. 18 Adversarial Distillation www.cnas.org
  19. 19 CISA and Partners Release Joint Advisory on Countering ... content.govdelivery.com
  20. 20 NSTM-4: US Policy Response to AI Model Distillation Attacks labs.cloudsecurityalliance.org
  21. 21 China State-Sponsored Cyber Threat: Advisories - CISA www.cisa.gov
  22. 22 Countering Chinese State-Sponsored Actors Compromise of ... - CISA www.cisa.gov
  23. 23 Alt CISO Daily Briefing — June 30, 2026 – Lab Space labs.cloudsecurityalliance.org
  24. 24 Examining the National Security Risks of PRC Artificial In www.congress.gov
  25. 25 China Bulletin: April 2, 2026 - uscc.gov www.uscc.gov
  26. 26 CISO Application Risk Intel Briefing for Week of September 2 www.veracode.com
  27. 27 Anthropic launches Opus 4.6 in another hit to the software ... finance.yahoo.com
  28. 28 Anthropic Economic Index report: Learning curves www.anthropic.com
  29. 29 Anthropic's Claude Opus 4.6 brings 1M token context and ' ... venturebeat.com
  30. 30 Anthropic's Transparency Hub www.anthropic.com
  31. 31 Review of the Anthropic Sabotage Risk Report metr.org
  32. 32 Distillation Attacks on AI Models by DeepSeek, Moonshot ... www.linkedin.com
  33. 33 Trump tech official accuses China’s Moonshot AI of stealing ... www.scmp.com
  34. 34 China AI labs accused of stealing from Anthropic's Claude ... www.foxnews.com
  35. 35 Anthropic Says DeepSeek Fraudulently Used Claude - Business ... www.businessinsider.com
  36. 36 Anthropic just dropped evidence that DeepSeek, Moonshot ... www.reddit.com
  37. 37 ANTHROPIC SAYS MOONSHOT USED 5,380 FRAUDULENT ACCOUNTS ... x.com
  38. 38 Anthropic Accuses Chinese AI Labs Of Industrial-scale ... dataconomy.com