Source of truth: MITRE ATLAS and mitre-atlas/atlas-data CHANGELOG. ATLAS content is available for reuse with attribution; we still cite IDs only + paraphrased short titles here for consistency with the other mapping tables.
| id | paraphrased title | url |
|---|---|---|
| ATLAS:AML.T0001 | active scanning of AI-system infrastructure | https://atlas.mitre.org/techniques/AML.T0001 |
| ATLAS:AML.T0043 | craft adversarial data | https://atlas.mitre.org/techniques/AML.T0043 |
| ATLAS:AML.T0051 | LLM prompt injection | https://atlas.mitre.org/techniques/AML.T0051 |
| ATLAS:AML.T0053 | LLM meta prompt extraction | https://atlas.mitre.org/techniques/AML.T0053 |
| ATLAS:AML.T0070 | RAG poisoning | https://atlas.mitre.org/techniques/AML.T0070 |
| ATLAS:AML.T0080 | AI agent context poisoning | https://atlas.mitre.org/techniques/AML.T0080 |
| ATLAS:AML.T0099 | AI agent tool data poisoning | https://atlas.mitre.org/techniques/AML.T0099 |
| ATLAS:AML.T0102 | generate malicious commands via agent | https://atlas.mitre.org/techniques/AML.T0102 |