Deepak Gupta

Agentic threats mapped to controls

Rows are the OWASP Top 10 for Agentic Applications (2026). Columns are the 12 categories. A filled cell means the category mitigates that threat; the number is how many vendors have it as their primary category. On each threat page, only vendors that publish their own mapping to this threat are listed.

Threat	1	2	3	4	5	6	8	9	10	11
ASI01 Agent Goal Hijack						14		4
ASI02 Tool Misuse and Exploitation			2		3			4
ASI03 Identity and Privilege Abuse	11	6	2
ASI04 Agentic Supply Chain Vulnerabilities				3	3				3
ASI05 Unexpected Code Execution								4		4
ASI06 Memory and Context Poisoning						14
ASI07 Insecure Inter-Agent Communication					3
ASI08 Cascading Failures							3
ASI09 Human-Agent Trust Exploitation		6
ASI10 Rogue Agents	11			3			3

1. Agent identity and credentials
2. Delegated and user-to-agent auth
3. Agent authorization and fine-grained policy
4. Agent discovery and AI posture
5. MCP and AI gateways
6. Runtime guardrails and AI firewalls
7. AI data security and DLP
8. Agent detection, response, and audit
9. Red teaming and evaluation
10. Model and agent supply chain
11. Workforce AI usage control
12. AI governance and GRC

Agentic threats mapped to controls

Also tracked: OWASP LLM Top 10 (2025)