AML.CS0059
EchoLeak: Zero-Click Prompt Injection Targeting M365 Copilot for Data Exfiltration
- Type
- Exercise
- When
- 2025-05
- Actor, as ATLAS names it
- Not stated
- Target
- Microsoft 365 Copilot
- ATLAS attributes it to Cato Networks (security evaluation).
- On ShieldMap: Microsoft 365 Copilot (matched by name in the target).
Techniques, in order
- 1LLM Prompt CraftingAI Attack Adaptation · AML.T0065
- 2Retrieval Content CraftingAI Attack Adaptation · AML.T0066
- 3Stage CapabilitiesResource Development · AML.T0079
- 4Prompt Infiltration via Public-Facing ApplicationInitial Access · AML.T0093
- 5LLM Prompt ObfuscationDefense Evasion · AML.T0068
- 6RAG PoisoningPersistence · AML.T0070
- 7LLM Prompt Injection: TriggeredExecution · AML.T0051.002
- 8Data from AI Services: RAG DatabasesCollection · AML.T0085.000
- 9LLM Trusted Output Components ManipulationDefense Evasion · AML.T0067
- 10LLM Response RenderingExfiltration · AML.T0077
- 11Exfiltration via Cyber MeansExfiltration · AML.T0025
- 12External HarmsImpact · AML.T0048
How these attacks work
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.