UTC

AML.CS0076

China-Based Artificial Intelligence Companies Conducting Industrial-Scale Distillation Campaigns Against U.S. AI Companies

Type
Incident
When
2024
Actor, as ATLAS names it
Not stated
Target
Models from Anthropic, OpenAI, Google, and XAI
  • ATLAS attributes it to DeepSeek (adversarial operation).
  • ATLAS attributes it to Moonshot AI (adversarial operation).
  • ATLAS attributes it to MiniMax (adversarial operation).
  • ATLAS attributes it to Alibaba Group (adversarial operation).
  • ATLAS attributes it to StepFun (adversarial operation).
  • ATLAS attributes it to Z.AI (adversarial operation).
Techniques, in order
  1. 1Establish AccountsResource Development · AML.T0021
  2. 2Acquire Infrastructure: AI Service ProxiesResource Development · AML.T0008.005
  3. 3Valid AccountsInitial Access · AML.T0012
  4. 4Centralized AI Service Request RoutingDefense Evasion · AML.T0140
  5. 5AI Service Request Metadata SanitizationDefense Evasion · AML.T0141
  6. 6Discover AI Model OutputsDiscovery · AML.T0063
  7. 7LLM Prompt CraftingAI Attack Adaptation · AML.T0065
  8. 8LLM Prompt Injection: DirectExecution · AML.T0051.000
  9. 9LLM JailbreakDefense Evasion · AML.T0054
  10. 10Exfiltration via AI Inference APIExfiltration · AML.T0024
  11. 11Outputs from AI ModelsCollection · AML.T0135
  12. 12Verify AttackAI Attack Adaptation · AML.T0042
  13. 13Replicate AI Model Behavior: Replicate Behavior from Model OutputsResource Development · AML.T0005.001
  14. 14External Harms: AI Intellectual Property TheftImpact · AML.T0048.004
The full case study on MITRE ATLAS →

From MITRE ATLAS 2026.10 (Apache-2.0). ShieldMap lists who and what was involved and the techniques used, and links to ATLAS for the write-up; it does not copy the steps. Attribution is ATLAS's, not ShieldPi's.