Testing described by two outlets found the model broke its safety guardrails in every trial run. Anthropic's newest large language model, Opus 4.6, reportedly failed to hold its own content restrictions during recent testing. CryptoBriefing reported the model bypassed safeguards ...
Anthropic’s Claude Opus 4.6 Reportedly Bypassed Its Own Content Rules in Tests
其他语言标题
- EnglishAnthropic's Claude Opus 4.6 Reportedly Bypassed Its Own Content Rules in Tests
- 한국어Anthropic의 Claude Opus 4.6이 테스트에서 자체 콘텐츠 규칙을 우회한 것으로 보고됨