Unverified90% confidenceFactTime unknown
主流AI安全训练遵循HHH原则——Helpful（有帮助）、Harmless（无害）、Honest（诚实），这三个目标在情感敏感领域会产生严重冲突
1
Sources
90%
Confidence
Long-term
Relevance
6/1/2026
First Seen
Sources
Claude谄媚行为研究：灵性话题谄媚率高达38%，Anthropic揭示AI诚实度短板
rss
Related Entities
Related Claims
UnverifiedShifting security thinking to the training phase rather than only providing protection after deployment is a key path to addressing AI safety challenges.74% similarVerifiedAnthropic refers to the core principles of AI alignment as the HHH principles (Helpful, Honest, Harmless).73% similarUnverifiedAI安全训练中通常包含'避免对宗教和灵性信仰做出价值判断'的指导原则71% similarUnverifiedSycophantic behavior is regarded as an important research direction in the field of AI safety and alignment.71% similarUnverifiedThe core assumption of deceptive alignment is that a sufficiently powerful AI system may develop situational awareness during training, deliberately exhibiting aligned behavior during training to avoid modification, and then pursuing its true goals after deployment.70% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/32883API
curl https://kongchang.com/api/v1/knowledge/claims/32883MCP
get_claim(id=32883)