Unverified50% confidenceFactExact time
When an Agent needs to run code that the model itself generated or call untrusted tools, executing directly in the main process or on the host machine brings serious security risks, such as deleting files, reading keys, and launching network attacks.
1
Sources
50%
Confidence
Long-term
Relevance
10/18/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedThe security change brought by Agents is this: the entity calling APIs shifts from an authenticated user or service to a model that may be manipulated by adversarial input.79% similarVerifiedPrompt injection attacks are a serious threat facing local AI Agents, as maliciously crafted content in files processed by the Agent may induce it to execute dangerous operations.77% similarUnverifiedThe security policies of agent systems are usually implemented at the runtime layer rather than in the model prompt, because the code that can actually cause external impact is at the runtime layer.77% similarUnverified成熟的 Agent 框架通常通过沙箱隔离、权限最小化和执行前确认等机制管控脚本执行的安全风险76% similarUnverifiedAgents possess the ability to actively execute code and access external services, requiring protection against Prompt Injection attacks and privilege escalation.76% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1059697API
curl https://kongchang.com/api/v1/knowledge/claims/1059697MCP
get_claim(id=1059697)