Why curbing chatbots' worst exploits is a game of whack-a-mole

AI companies are trying to impose safety measures on their chatbots, while researchers are finding ways around them all the time. Where will this end, asks Alex Wilkins