Unverified60% confidenceFactExact time
Anthropic uses the RSP framework to evaluate dangerous capability thresholds in models, OpenAI uses the Preparedness Framework, and Google DeepMind uses the Frontier Safety Framework
2
Sources
60%
Confidence
Medium-term (~90 days)
Relevance
8/7/2026
First Seen
Valid until: 11/5/2026
Sources
开放安全AI联盟成立：以开源筑牢AI安全防线
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73% similarUnverifiedOpenAI Preparedness团队的核心目标是在模型具备危险能力的早期阶段发现并评估这些能力72% similarUnverifiedThese safety frameworks stipulate that models must satisfy corresponding mitigation measures when they reach a specific risk level before they can be released.72% similarUnverifiedMajor vendors including Anthropic and OpenAI have successively established dedicated evaluation frameworks for cybersecurity abuse scenarios and incorporated test results into safety reports accompanying model releases.70% similarUnverifiedThe Preparedness Framework divides risk categories into four core domains: cybersecurity, chemical/biological/nuclear/radiological weapons (CBRN), model autonomy, and persuasion and deception70% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/699364API
curl https://kongchang.com/api/v1/knowledge/claims/699364MCP
get_claim(id=699364)