Unverified50% confidenceOpinionExact time
Models that can only generate text cannot break into systems, but agents with action capabilities can, with capability and risk growing in tandem.
1
Sources
50%
Confidence
Long-term
Relevance
9/23/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedAI is optimized to generate safe content that avoids errors, whereas true literary breakthroughs often come from breaking rules and defying expectations.68% similarUnverified由于大语言模型存在提示词注入攻击风险和幻觉问题，权限过大的Agent可能在被恶意输入诱导或判断失误时对业务数据、外部系统造成不可逆破坏66% similarUnverifiedLarge language models can only answer questions and cannot proactively initiate system calls or network requests, whereas agents can perceive needs, determine a course of action, and execute operations to complete end-to-end tasks.66% similarUnverified通过系统提示词约束智能体只回答知识库中有的问题，可有效防止智能体回答无关问题并引导转人工66% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/945089API
curl https://kongchang.com/api/v1/knowledge/claims/945089MCP
get_claim(id=945089)