Unverified50% confidenceSolutionExact time
给 AI 指令时应同时说明要做什么和不要做什么，主动划定边界以避免模型自由发挥
1
Sources
50%
Confidence
Long-term
Relevance
7/21/2026
First Seen
Sources
Related Claims
VerifiedAI model jailbreaking refers to using carefully crafted prompts or interaction patterns to bypass a model's built-in safety alignment mechanisms, causing it to produce restricted content.83% similarUnverified给 AI 指令时明确负向约束（Negative Constraints）与正向指令同等重要，能显著收窄模型输出的概率空间79% similarUnverifiedLarge models can only reason about which tool should be called, and actual execution still requires human intervention; whereas AI agents are capable of autonomously completing the full closed loop of tool invocation.79% similarUnverified当不懂操作时，向 AI 提问应问'你能不能自己干'而非'我应该怎么办'，因为前者会激活模型的工具调用能力自行执行79% similarUnverifiedPolicy enforcement for AI agents requires proactive interception, completing policy validation at the moment the AI issues a tool call request78% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/577481API
curl https://kongchang.com/api/v1/knowledge/claims/577481MCP
get_claim(id=577481)