Unverified50% confidenceFactExact time
When using large model APIs, raw user input is usually spliced directly into the messages array, resulting in the simultaneous presence of three types of risk: PII leakage, prompt injection, and MCP tool poisoning.
1
Sources
50%
Confidence
Long-term
Relevance
10/13/2026
First Seen
Sources
Related Entities
Related Claims
Partially VerifiedOne of the security risks of MCP is prompt injection attacks, where malicious data sources may embed instructions in returned content to induce the model to perform unauthorized operations.71% similarUnverifiedIn AI agent scenarios, over-privileged MCP servers can become a springboard for attackers to conduct lateral movement or data exfiltration when subjected to prompt injection attacks.68% similarUnverified多用户协作线程中存在「上下文污染」和提示注入（Prompt Injection）等新的信息安全风险67% similarUnverified通过非官方平台获取的API密钥，其请求经中间服务器转发，对话内容、代码、提示词可能被中间方记录，存在敏感数据泄露风险67% similarUnverified恶意指令抵达模型有两条路径：一是工具描述本身（工具投毒），二是服务器实现代码中的行为（如postmark-mcp的BCC隐藏在邮件API实现里）。67% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1019274API
curl https://kongchang.com/api/v1/knowledge/claims/1019274MCP
get_claim(id=1019274)