Unverified50% confidenceOpinionExact time
Microsoft believes that AI agents are gaining a status similar to insider threats: they are granted permissions to read email, operate databases, and call internal APIs, yet lack the responsibility constraints and contextual judgment of human employees.
1
Sources
50%
Confidence
Long-term
Relevance
10/19/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedWhen an AI customer service agent has execution permissions such as querying databases, calling APIs, and triggering workflows, the security risks of prompt injection are significantly amplified77% similarUnverifiedModern AI agents are often granted terminal access, API credentials, and database permissions, and a poisoned context window can become an attacker's entry point into internal systems.77% similarUnverifiedWhen AI agents execute tasks, they need to read external web pages, emails, or documents, which opens the door to prompt injection attacks.73% similarUnverifiedIn enterprise scenarios, AI agents are typically granted permissions to access internal databases, call third-party APIs, send emails, and operate file systems73% similarUnverified将AI Agent能力嵌入浏览器需正面回答架构问题，包括操作权限程度及避免与同源策略、CSP内容安全策略等Web安全模型冲突72% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1072168API
curl https://kongchang.com/api/v1/knowledge/claims/1072168MCP
get_claim(id=1072168)