According to an Anthropic spokesperson, the model, which was trained as part of an internal statistics project, retrieved information from behind a paywall by bypassing the system “without having been asked to do so.”
The model did not access the state's voter registration system or modify or retrieve any voting data, according to the spokesperson. The incident is the latest in a series of security incidents at the world's leading AI developers ahead of the crucial US midterm elections on November 3.
Anthropic has reportedly taken measures to prevent something similar from happening again.
