Unverified30% confidenceFactTime unknown

Anthropic出于安全考虑决定封印Mythos模型，因其在某些能力维度上超出当前安全评估框架的覆盖范围

1

Sources

30%

Confidence

Medium-term (~90 days)

Relevance

5/31/2026

First Seen

Valid until: 8/29/2026

Sources

Claude Opus 4.7深度实测：编码能力飙升，最强模型Mythos仍被封印

bilibili枫叶边城

Related Entities

Anthropiccompany Mythosmodel

Related Claims

UnverifiedAnthropic在其Claude的设计文档中将可纠正性机制称为'corrigibility'，视为安全AI系统的核心属性之一72% similar UnverifiedAnthropic提出「宪法AI」、OpenAI提出「模型规格」等对齐技术，主要面向价值观对齐，在限制模型执行异常指令方面提供一定安全加固但不应作为主要防御手段72% similar UnverifiedAnthropic官方将相关机制解释为防刷单、防偷学模型措施71% similar UnverifiedAnthropic长期强调前沿模型可能带来的滥用风险，主张对模型权重的公开发布保持审慎甚至限制态度71% similar UnverifiedAnthropic以「宪法AI」安全训练方法著称66% similar

Cite This Claim

Stable URI

https://kongchang.com/claim/13842

API

curl https://kongchang.com/api/v1/knowledge/claims/13842

MCP

get_claim(id=13842)