Anthropic 自 2022 年起预警 AI 网络攻击风险
Anthropic has been warning about AI-enabled cyberattacks at least since 2022, with @logangraham at the forefront of that work. Back then, no VC was taking it se
Anthropic 至少自 2022 年起就由 Logan Graham 牵头预警 AI 助力的网络攻击，但当时没有 VC 认真对待。Anthropic 还在 ChatGPT 相关真实事件发生前数年就构建了谄媚（sycophancy）评测，由 Ethan Perez 等人参与。作者认为，应在 AI 风险造成大规模伤害前就重视早期证据。
正文
Anthropic has been warning about AI-enabled cyberattacks at least since 2022, with @logangraham at the forefront of that work. Back then, no VC was taking it seriously. The paradox is that risks only feel real once they’re legible, but waiting for incidents is too late. Anthropic also built sycophancy evals (@EthanJPerez et al.) a couple years before real-world incidents with ChatGPT made the risks of models excessively agreeing with users legible at scale. The lesson is to take early evidence of AI risks seriously before harm at scale makes them impossible to ignore.