Unverified50% confidenceFactExact time
OpenAI的InstructGPT、Anthropic的Constitutional AI以及Meta的Llama系列都采用了类似的安全对齐流程
1
Sources
50%
Confidence
Long-term
Relevance
8/5/2026
First Seen
Sources
Related Claims
UnverifiedOpenAI和Anthropic等公司均设有专门的滥用检测系统77% similarUnverifiedOpenAI有Preparedness Framework，对模型在网络安全、CBRN威胁、说服力和自主行为四个维度进行分级评估76% similarUnverifiedOpenAI主要依赖RLHF（基于人类反馈的强化学习），而Anthropic的Constitutional AI代表方法论上的分野，追求安全原则的显式化与可解释性76% similarUnverifiedAnthropic主打宪法AI安全框架，训练分为SLAIF和RLAIF两阶段，与OpenAI采用的RLHF路径不同75% similarUnverifiedMeta的LlamaGuard和OpenAI的内容审核API采用类似的守门人分类器架构74% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/690530API
curl https://kongchang.com/api/v1/knowledge/claims/690530MCP
get_claim(id=690530)