Unverified60% confidenceFactExact time

OpenAI在2018年发布GPT-2时以能力过强可能被滥用为由分阶段限制模型权重公开

2

Sources

60%

Confidence

Long-term

Relevance

7/10/2026

First Seen

Sources

GLM-5.2引发媒体恐慌：开源AI的安全争议与治理困境

redditr/LocalLLaMA7/9/2026

Related Claims

UnverifiedGPT-2在发布时因生成能力过强，一度被OpenAI以被滥用风险为由延迟全量发布84% similar UnverifiedOpenAI 将 GPT-5.6 的发布节奏与五层安全闸门绑定：模型自身拒绝、实时分类器、账户维度审查、差异化权限、监控与停用78% similar VerifiedOpenAI在2018年发布了GPT-177% similar

Cite This Claim

Stable URI

https://kongchang.com/claim/456405

API

curl https://kongchang.com/api/v1/knowledge/claims/456405

MCP

get_claim(id=456405)