Unverified50% confidenceSolutionExact time
Tiered release divides model capabilities into different risk levels: architecture papers and evaluation benchmarks are fully public, medium-scale weights are made available to identity-verified research institutions, and the strongest version weights are only provided via a controlled API or not released externally.
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
9/15/2026
First Seen
Valid until: 12/14/2026
Sources
Related Entities
Related Claims
VerifiedOpenAI曾发布《准备框架》（Preparedness Framework），将风险分为低、中、高、关键四个等级，要求只有风险评估低于'高'等级的模型才能对外发布69% similarVerifiedOrganizations such as METR (Model Evaluation and Threat Research) are developing standardized dangerous capability evaluations, including whether a model can provide biosynthesis guidance beyond publicly available literature and whether it can autonomously discover zero-day vulnerabilities.67% similarUnverifiedA five-step framework for judging the credibility of AI model surpassing claims: whether evaluation data is publicly available, task coverage scope, whether the comparison baseline is reasonable, whether the publisher has a conflict of interest, and whether there is independent third-party replication.67% similarUnverifiedExecutive Order 14110要求开发者在发布具有双重用途风险的大型模型前，向商务部NIST、国土安全部等机构提交安全评估报告，并以训练所用浮点运算量（FLOPs）作为触发审查义务的量化阈值66% similarUnverified作者建议AI摘要系统对生死、健康、重大突发事件等高风险类别设置更严格核验门槛，为不同查询类型建立差异化信源可信度权重体系66% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/922336API
curl https://kongchang.com/api/v1/knowledge/claims/922336MCP
get_claim(id=922336)