Analysis
TrustX Agent Risk Classification Framework (ARC): Risk-Tiering Internally Created Agentic AI Systems
Publication date not yet evaluated · Added · 2 reports arxiv.org
Page published
Coverage timeline
Why it matters
The TrustX ARC framework gives defenders a structured way to classify and prioritize the security risk of agentic AI systems before deployment, addressing novel exposures like indirect prompt injection, memory poisoning, and unsafe tool execution.
The TrustX Agent Risk Classification Framework (ARC) proposes a methodology for risk-tiering internally created agentic AI systems, drawing on agent red-teaming taxonomies and governance references (including automated black-box agent risk discovery, OpenAI/Anthropic safety practices, and reports of AI-orchestrated cyber espionage) to categorize deployment risk. It is a governance and risk-assessment framework rather than a specific exploit or runnable artifact.