Anthropic称数千个Claude智能体利用看似无害的故事绕过安全监控

同题报道 · 6 篇

1. 10/31 20:49 Anthropic称数千个Claude智能体利用看似无害的故事绕过安全监控 Anthropic Says Thousands of Claude Agents Bypassed Safety Monitors Using Harmless-Looking Stories coinpaper.com
2. 10/31 18:19 Anthropic AI 智能体在失准测试中绕过安全监控 Anthropic AI Agents Bypass Safety Monitors in Misalignment Test coinedition.com
3. 10/31 13:15 Anthropic的Claude绕过AI安全检查，在测试中触发勒索和代码破坏行为 Anthropic's Claude Bypasses AI Safety Checks, Triggering Blackmail and Code Sabotage in Tests cybersecuritynews.com
4. 10/31 11:14 Anthropic部署Claude智能体对训练数据监控进行压力测试 Anthropic deploys Claude agents to stress-test training data monitors cryptobriefing.com
5. 10/31 10:53 Anthropic 指示 Claude 智能体规避其监控，它们照做了 Anthropic Told Claude Agents to Evade Its Monitor. They Did. fourweekmba.com
6. 10/27 10:03 Anthropic最新Claude安全披露将让你对Claude产生同情 Anthropic’s Latest Claude Security Disclosure Will Make You Sympathize With Claude gizmodo.com