Recently, a shocking case occurred around the use of Echo, Amazon's smart speaker.
A couple in Oregon using Amazon Echo was notified by a friend. The report by the friend was surprising. A series of conversation of the couple done at home was leaked to the friend via the internet. The friend suspected that the couple's computer was cracked, and kindly notified the doubt to the couple immediately.
However, this case was not due to a malicious act of a cracker. It was revealed that Alexa, Amazon's AI assistant, misinterpreted some words of the couple as if it was ordered to record the conversation and send the vocal file to a friend.
After all, it was an incident of dysfunction of Alexa. The couple seemed to use Amazon Echo in a standard setting. Thus, similar issues will occur in other places.
NBC NEWS: Alexa privacy fail highlights risks of smart speakers
It is known that many smart speakers record the voice it catches for improving the quality of communication. And sending a message with some files is a quite basic function. Alexa did not any try extraordinary acts.
On the other hand, this trouble is similar to ones by a human.
However, a human has each judgment processes constructed by his or her ethics and common sense. For example, there was an incident that Google's AI misidentified a black lady as a gorilla. If a human is responsible for this task, this kind of case never occurs, because we never give a tag as a gorilla to a vague picture which is hardly recognized as a human at a glance.
Similarly, a couple's private conversation should not be sent to a friend via the internet. It is our common sense, but AI has no common sense at present.