Taking you to the card → For once, the AI safety news might be very good