Unverified50% confidenceFactExact time
Prompt injection attacks can be divided into two categories: direct prompt injection and indirect prompt injection, with indirect prompt injection being more harmful because the attack surface encompasses all external data sources reachable by the agent.
1
Sources
50%
Confidence
Long-term
Relevance
9/12/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedJailbreak attacks refer to bypassing a model's safety alignment mechanisms through carefully crafted prompts to make it output harmful content.78% similarUnverified自动化对抗样本生成面临覆盖率幻觉挑战，自动生成的攻击样本可能数量庞大但高度同质化，遗漏新颖攻击路径71% similarUnverifiedfixture作用域越大，被共享范围越广，引入隐性依赖的风险也越高69% similarUnverifiedBreaking down vague concerns about AI loss of control into specific risk categories enables more targeted preventive measures.68% similarUnverifiedSandbox isolation design can limit the impact scope when an agent encounters a malicious prompt injection attack68% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/916048API
curl https://kongchang.com/api/v1/knowledge/claims/916048MCP
get_claim(id=916048)