Voice assistants are increasingly popular, accompanied by concerns about privacy and accuracy. The Guardian said Apple hired many researchers to listen to user conversations recorded by Siri.
They are tasked with scoring Siri’s answers based on many factors, including how accurate this virtual assistant works. This is aimed at improving product quality. One of the researchers revealed that the recordings contain a lot of private information.
This virtual assistant is often mistakenly activated and recorded immediately afterwards. The keyword “Hey Siri” can activate virtual assistants with homophone words, even the sound of locking clothes. The Apple Watch also starts this function automatically when users turn their wrists to watch the time.
Many random recordings revealed users’ personal lives such as doctor appointments, illegal business agreements, and sensitive sounds while “rain clouds”. The listening sections also provide location, contact information and application data.
 |
| Siri is often mistakenly activated and recorded immediately afterwards. Photo: Phonehouse . |
Apple’s privacy page has written, to help virtual assistants identify their voices and provide better feedback, some user data will be encrypted and sent to Apple servers for research. However, no line mentioned that data will be heard and analyzed by real people.
Apple has repeatedly criticized Google and Facebook for not respecting the privacy of users, but their work is no different from those firms.
Earlier this year, according to Bloomberg , Amazon took unauthorized user data. Instead of collecting voice recordings, the Alexa assistant integrated into the company’s Echo smartphone has turned them into text and sent them to the remote server.
In mid-July, The Verge reported VRT NWS (Belgium) TV channel accused Google of paying research parties to get the audio content that virtual assistant Google Assistant obtained from users.
 |
The Cupertino giant confirmed that the data they used was less than 1% of the information the virtual assistant obtained and was not linked to the user account, but that contained a lot of sensitive information from users.
The Guardian insists that, while other companies allow users to deny the use of some of their recordings, Apple does not have this option.
This incident raises a series of questions: What will Apple and others do to protect users’ privacy when they develop virtual assistants? Should users be notified when companies use their recordings? What can they do to reduce accidental triggers? How should companies process information from research parties? Who will be responsible when related issues occur?
Virtual assistants are a good product, and also a warning bell for future technologies to be carefully considered for consequences before they are developed and used.