Anthropic has dropped a controversial new AI disclosure that, at first glance, feels both remarkable and unnerving.

Perfect AI alignment with human values and interests is mathematically impossible, according to a study, but behavioral ...