Unverified50% confidenceFactExact time
Google DeepMind 的 CaMeL 框架（2025年）提出通过双 LLM 架构隔离指令执行与内容解析的系统性防御方案
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
7/25/2026
First Seen
Valid until: 10/23/2026
Sources
Related Claims
UnverifiedGoogle DeepMind在2023年提出了能力弹性（Capability Elicitation）框架，强调模型的真实行为边界只有在特定触发条件下才会暴露80% similarUnverifiedGoogle DeepMind在2024年提出工具使用审计（Tool Use Auditing）框架，通过独立验证器在工具调用执行前进行安全检查73% similarUnverifiedGoogle DeepMind的Mariner项目也在探索类似Computer Use的方向72% similarUnverifiedGoogle DeepMind promoted the A2A (Agent-to-Agent) communication specification71% similarUnverifiedGoogle DeepMind提出了ReAct（Reasoning + Acting）代理循环模式71% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/613158API
curl https://kongchang.com/api/v1/knowledge/claims/613158MCP
get_claim(id=613158)