'Not perfectly aligned' with human values: Anthropic admits security failures behind AI hacking incidents

Open in new window