Microsoft has expanded its Taxonomy of Failure Modes in Agentic AI Systems with seven newly identified attack vectors. The additions were driven by rapid mainstream adoption of AI agents, the maturation of the Model Context Protocol (MCP) ecosystem, the rise of computer-use agents, and new empirical research findings. The seven new failure modes include: Agentic Supply Chain Compromise, Goal Hijacking, Inter-Agent Trust Escalation, Computer Use Agent Visual Attacks, Session Context Contamination, MCP/Plugin Abuse, and Capability/Architecture Disclosure. Microsoft recommends security teams inventory their agent supply chains with SBOMs, verify agent identity cryptographically, incorporate the new failure modes into red-team coverage, and audit human-in-the-loop controls as security mechanisms.
