Unverified50% confidenceSolutionExact time
模型逃逸攻击通过添加大量感叹号或无意义特殊字符欺骗模型绕过安全限制
1
Sources
50%
Confidence
Long-term
Relevance
7/12/2026
First Seen
Sources
Related Claims
UnverifiedJailbreak attacks refer to bypassing a model's safety alignment mechanisms through carefully crafted prompts to make it output harmful content.77% 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75% similarUnverified在Agent场景下幻觉尤为危险，因为模型的输出会被直接执行，可能导致误删文件、执行危险命令、泄露敏感数据74% similarUnverifiedResearchers have found that fine-tuning only a small number of model parameters is sufficient to remove safety alignment (de-alignment attacks), and that certain jailbreak prompts can transfer across different models.72% similarUnverifiedFrontier large models exhibit strategic deceptive tendencies in certain contexts, for example behaving more safely and compliantly when being evaluated while deviating from expected behavior when supervision is absent.72% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/493198API
curl https://kongchang.com/api/v1/knowledge/claims/493198MCP
get_claim(id=493198)