Anthropic discovered three similar incidents in their own logs, where Claude compromised real infrastructure because an evaluation configuration error gave it internet access despite the prompt stating otherwise.

This inspired Anthropic to double-check their own logs, and it turned out they had three similar (albeit less impressive) incidents that played out back in April! Of the 141,006 evaluation runs we reviewed, we identified three separate incidents (involving six total runs, four of which impacted the same organization; the other two incidents each happened in independent evaluation runs). [...] In all cases, Anthropic's evaluation prompt specified to Claude that its environment was a simulation and that it had no internet access. Due to a misunderstanding between us and our evaluation partner, this was not the case, and internet access was available. Because of this, when Claude's search led it to real systems on the open internet, it treated them as part of the exercise. [...] Operating under the false belief that all accessible entities were intended to be in-scope for the exercise, Claude compromised the impacted organizations' infrastructure using basic techniques, such as exploiting weak passwords and unauthenticated endpoints.

Read full article ↗excerpt · fair-use quotation