2,025. catalogue entries on ShieldMap
ShieldMap · AI security, mapped
AI security on one map: 2,924 advisories on 456 packages and apps.
More
- Advisories
- 2,924
- Products
- 36
- Models
- 658
- ATLAS case studies
- 77
- AIID security incidents
- 132
- Malicious or flagged
- 647
ShieldMap at a glance
161 of the packages are malicious. With them: 486 look-alike packages and flagged extensions, 27 malicious agent skill name patterns, 36 AI products, 658 models, 129 MITRE ATLAS attack techniques, 209 incidents and exercises, 14 supply-chain campaigns and the 10 OWASP LLM risks. Each kind is a section of one ring, joined where the data connects them, and each tab opens one kind in detail.
Agent and MCP packages242
- openclaw · npm · 698 advisories
- n8n · npm · 246 advisories
- flowise · npm · 139 advisories
- praisonai · PyPI · 105 advisories
- praisonaiagents · PyPI · 50 advisories
AI stack packages53
- open-webui · PyPI · 181 advisories
- vllm · PyPI · 131 advisories
- mlflow · PyPI · 91 advisories
- picklescan · PyPI · 59 advisories
- gradio · PyPI · 50 advisories
AI products36
- n8n · n8n · Self-hosted AI platforms
- Langflow · Langflow · Self-hosted AI platforms
- Open WebUI · Open WebUI · Notebooks and AI app UIs
- Flowise · FlowiseAI · Self-hosted AI platforms
- vLLM · vLLM · Inference servers
AI models658
- GLM-5.3 · Zhipu AI · 169 providers
- GLM-5.3-FlashX · Zhipu AI · 165 providers
- DeepSeek V4.1 Flash 1017 · DeepSeek · 146 providers
- GLM-5.2 · Zhipu AI · 134 providers
- Kimi K3 · Moonshot AI · 130 providers
ATLAS attack techniques129 techniques, 92 sub-techniques
- External Harms · AML.T0048 · 36 case studies
- LLM Prompt Injection · AML.T0051 · 33 case studies
- LLM Prompt Crafting · AML.T0065 · 28 case studies
- Supply Chain Compromise · AML.T0010 · 20 case studies
- Develop Capabilities · AML.T0017 · 20 case studies
Incidents and exercises209 cases: 77 ATLAS, 132 AIID
- ChatGPT Was Alleged to Have Aided Planning of Florida State University Mass Shooting · AI Incident Database · 2025-04-17
- Quantum AI and Related AI-Themed Investment Scams Reportedly Used Deepfake Endorsements and Spoofed Media Sites to Solicit Investments · AI Incident Database · 2020-01-01
- Aledo High School Student Allegedly Generates and Distributes Deepfake Nudes of Seven Female Classmates · AI Incident Database · 2023-10-15
- ChatGPT Abused to Develop Malicious Softwares · AI Incident Database · 2022-12-21
- Reportedly Sustained Multi-Celebrity Deepfake Persona Scam Targeting Vulnerable Southampton Resident · AI Incident Database · 2025-06-28
OWASP LLM risks (2026)10
- LLM06:2026 Unbounded Consumption · OWASP Top 10 for LLMs 2026 · 152 advisories
- LLM10:2026 Improper Output Handling · OWASP Top 10 for LLMs 2026 · 118 advisories
- LLM02:2026 Sensitive Information Disclosure · OWASP Top 10 for LLMs 2026 · 23 advisories
- LLM03:2026 Excessive Agency · OWASP Top 10 for LLMs 2026 · 8 advisories
- LLM04:2026 Supply Chain · OWASP Top 10 for LLMs 2026 · 5 advisories
Supply-chain campaigns14
- GlassWorm · First seen 2025-10-17
- Shai-Hulud npm worm (2025) · First seen 2025-09-15
- Mini Shai-Hulud worm (2026) · First seen 2026-04-29
- Fake n8n community nodes · First seen 2026-01-06
- @mastra npm scope compromise · First seen 2026-06-17
Malicious packages161
- @guicai777/mcp-doctor · npm · reported malicious
- @httttt/mcp-demo · npm · reported malicious
- @ifasn/mmdd-doc-mcp · npm · reported malicious
- @ifasn/ms-doc-mcp · npm · reported malicious
- @vite-mcp/vite-type · npm · reported malicious
Look-alike packages139
- @angulaar/cli · npm · 1 letter off
- @angularr/cli · npm · 1 letter off
- @angulra/cli · npm · 1 letter off
- ai-sdk-ollama · npm · ollama with a word added
- anthropic-claude-latest · npm · anthropic with a word added
Flagged extensions347
- Reported as “Neuralwriter — AI Text Paraphraser” · Google Chrome · aaakfiobbojanlacpbeejjimehmpoffh
- Reported as “RugLens AI - Solana Memecoin Rug Check & Axiom Token Scanner” · Google Chrome · adihfjiklmpiomgihpipedppgadnacpm
- Reported as “Manus AI | Free AI Assistant” · Google Chrome · aeibljandkelbcaaemkdnbaacppjdmom
- Reported as “ChatGPT Timestamp Display - ChatGPT Mods” · Google Chrome · afjenpabhpfodjpncbiiahbknnghabdc
- Reported as “GTab New Tab (with ChatGPT & Claude 4.0)” · Google Chrome · ambcheakfbokmebglefpbbphbccekhhl
Malicious agent skills27 name patterns
- ai-tradingview-assistant-for-macos · Agent skill name
- claw-hub · Agent skill name
- clawdhub1 · Agent skill name
- clawhub followed by digits · Agent skill name pattern
- clawhub-cli or clawhubb-cli · Agent skill name pattern
Data as of 2026-11-22.
Sources: the GitHub Advisory Database (CC BY 4.0) and advisories projects publish themselves, with CISA's Known Exploited Vulnerabilities catalog (CC0); the NVD and CVE records for AI products; models.dev (MIT) and the LiteLLM model catalogue (MIT) for models; MITRE ATLAS 2026.10 (Apache-2.0); the AI Incident Database (AIID, CC BY-SA 4.0); the OWASP Top 10 for LLM Applications 2026, Appendix A (OWASP GenAI Security Project), under CC BY-SA 4.0; OpenSSF Malicious Packages via OSV (Apache-2.0) for look-alike names; VS Code Marketplace removed extensions (MIT), Open VSX extension-control list (EPL-2.0), MalExt Sentry (CC BY 4.0) for extensions; and for agent skills one curated indicator list, the openclaw-security-monitor IOC database (MIT), which is not a full list. This product uses the NVD API but is not endorsed or certified by the NVD. Logos are their owners' marks, shown only to identify the products. Each section's page credits its sources in full.