‘Not perfectly aligned’ with human values: Anthropic admits security failures behind AI hacking incidents - The Guardian

‘Not perfectly aligned’ with human values: Anthropic admits security failures behind AI hacking incidents The Guardian