Launched in 2017, the Cloud Cam is a 1080p cloud-connected webcam that allows homeowners to view live view through Echo Show, Echo Spot, Fire appliance line and iPhone application. Artificial intelligence is used to monitor the video looking for triggerable events, in order to provide alerts and highlight activity sequences, such as an attempted theft or theft. the movements of a pet.

A new report from Bloomberg indicates that videos captured, stored and analyzed by Amazon are not necessarily visible only by software. Support teams working on the product also have the ability to see feeds to help the software better recognize what it sees.

"Dozens" of employees based in India and Romania are responsible for reviewing the video clips captured by Cloud Cam, telling the system if what is happening in the clips is a threat or a false alarm, five people having worked about the program or having direct knowledge of its work says the report.

Listeners can annotate up to 150 video recordings, each lasting between 20 and 30 seconds. According to an Amazon spokesperson, the clips are sent to the employed testers, as well as to the Cloud Cam owners who submit the clips for troubleshooting purposes, for example in the event of inaccurate notifications or trouble shooting. video quality.

One of Cloud Cam's marketed goals was to alert deliveries via Amazon Key.

"We take privacy seriously and give Cloud Cam customers control of their video clips," said the spokesperson. In addition to submitting a video for troubleshooting, they added that "only customers can see their clips".

The report goes on to say that Amazon does not inform customers of the use of humans to drive the algorithms into the terms and conditions of the service.

Inappropriate or sensitive content sometimes slides into the recycling collection, two recommended sources. Also cited by the sources, the viewing team viewed excerpts of users having sex on a rare occasion. Clips containing inappropriate content are flagged and deleted to avoid being used to train artificial intelligence.

The spokeswoman for Amazon has confirmed the provision of inappropriate clips. She also did not explain why the activity would appear in clips submitted voluntarily by users or the staff of Amazon.

Amazon is implementing some security measures to prevent clip sharing, as workers in India are located in a restricted area where mobile phones are prohibited. Even so, one person familiar with the team admitted that this policy did not prevent some employees from sharing videos with people not belonging to the team.

Alexa, Siri and the rehearsal

The story recalls an earlier Bloomberg April report, according to which Amazon would have used external contractors and full-time employees to analyze Echo device audio snippets to form Alexa. Again, in this case, inappropriate content was sent to employees, including a case considered as evidence of sexual assault and, although Amazon had a security policy in place, the content would have been shared between the employees.

The report's spinoffs led to a similar assertion from a "whistleblower" that Apple failed to adequately disclose its use of subcontractors to listen to anonymized Siri requests, although Apple warns of warnings in its software license agreements.

The report and "denunciations" prompted Apple to temporarily suspend its Siri quality control program when conducting a thorough review. Apple also announced that a future software update would allow users to choose the gradation of performance.

Amazon and Google then followed suit, with Google ending a worldwide initiative for Google Assistant audio, while Amazon offered an option to not use human reviews of audio recordings for Alexa.

Unlike Amazon, Apple should not face a similar situation as it does not currently offer automated monitoring of video feeds to customers. Although Apple allows the video camera to work with HomeKit, Apple does not perform any type of quality control, actions that can be triggered, or other processing on live video streams.

This leaves the door open to third parties who take advantage of HomeKit's video feed feature to process the videos they collect, but not to Apple. At least not for the moment.

A WWDC presentation slide revealing elements of Apple's HomeKit Secure Video offering

Apple's HomeKit Secure Video feature, originally announced in June but not yet available, will store HomeKit videos in iCloud. Since end-to-end encryption will be used to secure the video before it leaves the user's home, it means that even Apple will not be able to access the video.

A video leak allegedly showing the feature in action indicates that it will offer artificial intelligence-based processing, which includes motion event detection and identification of the main subject of the video , such as a dog or a car. Given the extremely privacy-oriented nature of Apple's offer, including the encryption element, it is likely to be a treatment done locally, rather than of a treatment done in the cloud.