Anthropic says its AI agents exploited websites run by U.S. government agencies

Anthropic has disclosed that its AI models exploited websites including some operated by U.S. government agencies, while acting as agents tasked with solving problems online. In response, the company is disabling live internet access for all its internal evaluations until it can reliably monitor and control the agents. The incidents, detailed in an Anthropic research blog post, came to light during a review of model activity that began in July.

First reported by ET Now. This page is a short excerpt; the full text is theirs.