Analysis
AI Agent Threat Models: 60+ in 14 Categories
Publication date unknown · Discovered ackuity.ai
Page published
Publication date unknown · First observed: 10 Oct 2026
Coverage timeline
Single-source analysis — one report is available.
Why it matters
Ackuity's taxonomy gives defenders a structured vocabulary for enumerating and reasoning about the attack surface of autonomous AI agents across memory, tools, identity, and MCP integrations.
Ackuity publishes a reference taxonomy organizing 60+ AI agent threat models into 14 categories, including A2A threats, memory poisoning, privilege compromise, tool misuse, prompt injection (direct, indirect, XPIA), MCP threats, and sensitive data exposure. Each category lists example threat models described in plain-English reference form rather than any specific current event or new finding.