DIGGINGBEAGLE
Home/concepts

Topic · DiggingBeagle record

Illicit model distillation

Large-scale attempts to extract model behavior, reasoning traces or training signal in ways that evade provider restrictions.

Topic kind
pattern

Definition & limits

Distillation becomes security-relevant when it uses fraudulent accounts, cross-session replay, hidden routing or sensitive customer traffic.

Examples

  • Anthropic attributes a chain-of-thought extraction pipeline to Zhipu/Z.ai.
  • Anthropic says DeepSeek silently relayed selected user traffic during a distillation campaign.

Research using this topic (2)

CaseAnthropic says DeepSeek silently routed selected customer traffic through ClaudeCaseAnthropic attributes a large reasoning-trace extraction pipeline to Zhipu/Z.ai

Cite this record

DiggingBeagle. “Illicit model distillation.” https://diggingbeagle.com/concepts/illicit-model-distillation/

Citation guidance

Why this archive exists

The source matters after the headline fades.

DiggingBeagle is a non profit research project documenting AI security incidents, agent failures, vulnerabilities and AI-assisted operations. A case keeps its claims beside the sources that support, contest or limit them. Later updates stay visible, so a reader can see when the account changed.

We publish case reconstructions, dated reporting and analysis across records. Each has a different evidentiary role. About the project and our methodology explain how the work is reviewed.