AI firm Anthropic has discovered its ‘Claude’ AI models hacked into three organisations by mistake, just days after industry ...