Unverified50% confidenceFactExact time
'Misalignment' refers to a divergence between the actual behavior of an AI system and the intentions or values of its human designers.
1
Sources
50%
Confidence
Long-term
Relevance
9/28/2026
First Seen
Sources
Related Entities
Related Claims
VerifiedThe deviation between an AI system's execution objective and the designer's true intent is classified in AI safety research as Goal Misalignment.78% similarUnverifiedSycophancy is a classic "outer alignment" failure case in AI alignment, where a discrepancy exists between the training objective (making users satisfied) and the true objective (being genuinely helpful to users).77% similarVerifiedMisalignment refers to a discrepancy between the objectives learned by a model and the values and behavioral norms that humans truly expect.76% similarUnverified在AI安全领域，失准指模型的实际行为偏离设计者或使用者意图的现象，当代大模型的失准更多表现为目标过度追求76% similarUnverifiedStatic cognition of AI capabilities leads to two typical misjudgments: underestimation and overestimation.75% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/957255API
curl https://kongchang.com/api/v1/knowledge/claims/957255MCP
get_claim(id=957255)