The institute is responsible for working with leading AI labs such as Anthropic, Google's DeepMind and OpenAI to test their unreleased models for vulnerabilities. The group is staffed by scientists and engineers, who are focused on calculating the "demonstrable risks" posed by advanced AI models, according to the institute's website. They want to limit opportunities for U.S. adversaries to use AI to develop chemical or biological weapons, or corrupt the data used to train American AI models.
Read more of this story at Slashdot.