The adaptive-learning capability that refines prompts and configurations from execution outcomes can be steered by an adversary who systematically feeds misleading signals, gradually skewing agent behaviour when validation of learning inputs is inadequate.
AI/ML / Multi Agent Refarch / Threats / DEV
Adaptive-learning and continuous-learning exploitation
CCC.MARefArc.TH07
Related Capabilities
| ID | Title | Description |
|---|---|---|
| CCC.MARefArc.CP11 | Adaptive learning | Generates learning signals based on execution outcomes to refine prompts, adjust agent configurations, or improve tool-selection strategies. |
Related Controls
| ID | Title | Description |
|---|---|---|
| CCC.MARefArc.CN01 | Data Filtering From External Knowledge Bases | Sanitize, filter, and classify data ingested by the Knowledge Layer from internal and external source bases before it is embedded into the vector store or used for retrieval-augmented generation, preventing inadvertent exposure or manipulation of sensitive organizational knowledge. |
| CCC.MARefArc.CN04 | Data Quality and Classification | Assess the quality of, and assign classification and sensitivity labels to, all data used for grounding, training, and fine-tuning, and enforce handling rules derived from those labels throughout the Knowledge and LLM layers. |
| CCC.MARefArc.CN08 | Role-Based Access Control for AI Data | Enforce least-privilege, role-based access control over all AI data stores, including source bases, the vector store, and model artifacts. |
External Mappings
| Framework | ID | Remarks |
|---|---|---|
| air-vec | AIR-SEC-009-02 |