Unverified50% confidenceOpinionExact time
General-purpose large language models exhibit inconsistency in security judgments, and the same risky operation may bypass their security mechanisms simply by being rephrased or provided with a justifying context
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
9/11/2026
First Seen
Valid until: 12/10/2026
Sources
Related Entities
Related Claims
Unverified安全对齐本质是在模型概率分布上施加偏置，而非从根本上消除有害能力80% 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.78% similarUnverifiedOver-refusal strategies reduce risk but harm model utility, especially in scenarios such as professional development and security research that require touching sensitive boundaries.77% similarUnverified通用越狱指单一提示词或操作序列能系统性绕过模型所有安全约束，而针对性越狱只能解锁某一类具体有害行为76% similarUnverified不同厂商在RLHF阶段对安全性与有用性的权重取舍不同，过于强调规避风险会使模型给出保守、模糊、附带免责声明的回答75% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/904820API
curl https://kongchang.com/api/v1/knowledge/claims/904820MCP
get_claim(id=904820)