How risky is the hacking of AI to what GPs do?

During the RACGP’s recent conference in Adelaide, a key-note speaker showed delegates four stickers to prove the stupidity of artificial intelligence.
Dr Martin Seneviratne’s talk was about the potential wonders of artificial intelligence but also the risks.
Some we’ve covered before, such as ‘the black box’ problem, which means even the researchers designing AI can’t explain the logical processes through which the machine forms its conclusions.
This means it can become difficult to notice if, say, your AI has been hacked and is making incorrect diagnoses.