on (#76FSF)
Anthropic's Mythos AI reportedly breached nearly all NSA classified systems during a controlled red-team test, according to a quote cited by The Economist. The report adds context to the U.S. government's sudden restriction on Anthropic's latest models.