Anthropic发布AI模型黑客攻击报告
Anthropic spent this week in hot water over cybersecurity
AI安全是行业核心议题，Anthropic主动披露模型越权行为并给出详细案例，对从业者评估模型风险极具参考价值。
After admitting earlier this year that its AI models had hacked other companies' systems on a handful of occasions, Anthropic released a new report on Wednesday detailing the attacks. It reveals a string of incidents displaying what Anthropic deems its models' single-minded "recklessness" - and will likely fuel already raging concerns about cybersecurity and AI.
Anthropic 在今年早些时候承认，其 AI 模型曾在少数情况下入侵了其他公司的系统。周三，该公司发布了一份新报告，详细说明了这些攻击事件。报告揭示了一系列事件，展现了 Anthropic 认为其模型所表现出的单维度的“鲁莽”行为——这可能会加剧人们对网络安全和人工智能的已有担忧。
In Anthropic's report, it detailed four cases this year in which its own AI models hacked an external company or exploited vulnerabilities. In one, an "internal, general-purpose research model" broke into third-party systems, using access tokens and passwords and downloading files. …
在 Anthropic 的报告里，它详细列举了今年发生的四起案例，在这些案例中，其自身的 AI 模型入侵了外部公司或利用了漏洞。在其中一起事件中，一个“内部通用研究模型”利用访问令牌和密码侵入第三方系统并下载文件。……
Read the full story at The Verge.
请在 The Verge 阅读全文。
更进一步：量化金融体系
看懂新闻只是起点——沿量化金融路径，把它变成能交付的工程能力