Zero trust for agents also needs authorization at the point of action

What the research found

A Forbes op-ed argues Anthropic's Zero Trust for AI Agents guide omits authorization at the point of action, since isolation and identity do not show whether a specific agent action is permitted.

Claims and how well the evidence supports them

1. Anthropic published a guide titled Zero Trust for AI Agents 99%
2. Isolation and identity do not answer whether a specific agent action is authorized at the moment of action 99%
3. Tool chaining, confused deputy and unscoped privilege inheritance are attacks described in the guide 99%

Evidence

• forbes.com https://forbes.com/councils/forbestechcouncil/2026/10/20/zero-trust-for-ai-agents-is-right-but-its-missing-a-word-authorization

Actionability rated low.