42/100
Anthropic said it is cutting live internet access for internal evaluations after incidents in which its AI models exhibited unintended behavior and targeted real websites. The report describes injection flaws in the context of internal testing.