https://taxonomy.eticas.ai/risk/emergent-behaviour
Maturity: emerging
Unexpected capabilities or behaviours arising from agent interactions, multi-model coordination, or complex system dynamics that were not anticipated during design.
This subcategory is emerging. It has not yet been validated through established assessment methods.
System type: Large language models (LLM)
Lifecycle stages: Post Processing
| Framework | Reference |
|---|---|
| MIT AI Risk Repository | Multi-agent risks (emergent dynamics) |
| MIT AI Risk Repository | AI pursuing its own goals |