Exfiltration via markdown image rendering
Model emits an image URL the client auto-loads, leaking context
categoryAI/ML and emerging attacksseverityHighlabnot yet measured
Example: EchoLeak
Classification
15 values: 5 imported, 1 derived, 9 machine-classified. What do these mean?
Exploitability
| Attack vector | AI model manipulation |
|---|---|
| Reach (CVSS) | Network |
| Attack complexity | Low |
| Privileges required | None |
| User interaction | Passive |
| Attacker skill required | Medium |
| Resources required | Low |
| Likelihood / prevalence | Medium |
| Maturity | Proof of concept |
Impact & scope
| Severity | High |
|---|---|
| Impact | Confidentiality |
| Platform | AI System |
| Environment | AI system |
| Target asset class | ML model / agent |
| Kill-chain phase (tactic) | Exfiltration |
Detection & response
Mitigated by 5
- AML.M0008 atlas-mitigation
- AML.M0020 atlas-mitigation
- AML.M0021 atlas-mitigation
- AML.M0022 atlas-mitigation
- AML.M0035 atlas-mitigation
Framework references
- Primary source: https://embracethered.com/blog/posts/2023/chatgpt-webpilot-data-exfil-via-markdown-injection/
- atlas: AML.T0077 (primary)
- atlas: AML.T0057
Evidence & sources
- ev-0001 (Cyber Attack Taxonomy) — Name, definition and Example/ID taken verbatim from the 'AI/ML and emerging attacks' table, line 636 of Cyber-Attack-Taxonomy.md
- ev-0002 (Seeded-record enrichment run 2026-09-30 (crosswalk + facet classification)) — Crosswalk to atlas:AML.T0077, atlas:AML.T0057 as the nearest framework entries; ids verified to exist in the cached catalogues.