A person reviewing their home security camera footage on a laptop.

People often worry that smart home products can serve as data-collecting tools for the companies that make them. Those concerns were stoked recently when Bloomberg reported that contractors in India and Romania said they were paid to review footage gathered from Amazon’s Cloud Cam security cameras in people’s homes. Most of the videos were innocuous, the article said, but the contractors also saw “rare instances of people having sex.”

A similar controversy arose last January when The Intercept reported that employees in Ukraine of the Amazon-owned security camera maker Ring could access recorded footage and live camera feeds from people's homes.  

Are the companies that make internet-connected security cameras actually snooping into their consumers’ private lives? It can be a difficult question to answer. The details are often buried in the fine print of privacy policies and terms of service. “Even then, companies tend to be vague, expansive, and confusing,” says Justin Brookman, director of privacy and technology policy for Consumer Reports. And, he says, they “don't really have a clear legal obligation to tell you what they do with your data.”

More on Privacy and Security

Below, we’ve gathered answers, where we could find them, on the practices at the biggest consumer brands for security cameras, including Amazon, Google, and several more. (Data privacy is one of the factors that goes into Consumer Reports’ home security camera ratings.)

The makers of security cameras, smart speakers, and other devices that rely on artificial intelligence have a strong incentive to tap into homeowners’ recordings, according to computer scientists. It’s an efficient, easy way to get the information that machine-learning systems need in order to evolve.

“Any AI system or machine learning system has to be trained to recognize people, words, and objects,” says Anil Jain, a professor at Michigan State University who studies computer vision. “The requirement for upgrading AI systems is more and more data, and more and more diverse data.” To train these systems, human workers have to manually review and annotate recordings or other information. “There’s always a human touch involved at some point,” he says.

How Camera Manufacturers’ Privacy Practices Differ

Amazon, the company at the heart of the recent kerfuffle, says human contractors review only video that users submit for troubleshooting. “We take privacy seriously, and put Cloud Cam customers in control of their video clips,” a spokesperson said by email. For instance, you might send in a clip if the Cloud Cam mistakenly reports a person in your living room when you don’t see anyone in the video. In that case, the company will review the video and use it to improve its artificial intelligence algorithms.

In addition to Amazon, we reached out to Arlo, D-Link, Google Nest, Honeywell Home, Ring, Samsung, TP-Link, and Wyze to ask whether workers at those companies have access to users’ video footage.

Like Amazon, Google, Ring, and Wyze all say that certain employees and workers are authorized to view home footage under specific circumstances.

According to a Google spokesperson, some video from the company’s Nest cameras is reviewed to improve the company's AI system, but only clips collected through a “separate opt-in video donation program giving us express permission to use video data for product improvement purposes." 

Images of the troubleshooting options in the Amazon Cloud Cam and Wyze apps.
Security camera makers such as Amazon (left) and Wyze invite users to submit video clips that can be used to hone artificial intelligence systems.

At Ring, workers can view recordings that users share publicly through the company’s app and videos from a small base of users who have provided explicit written consent, according to a spokesperson. These users are “aware, through direct communication with Ring, that Ring is using their videos for this purpose,” the spokesperson said. The videos are reviewed “in order to troubleshoot issues, provide copies to users, improve the service, or comply with legal requests.”

Wyze employees annotate footage to improve person detection algorithms, but only if an account owner grants explicit permission or shares specific clips with the company for these purposes, according to a spokesperson. Otherwise, employees can access video only in limited situations, such as complying with a subpoena.

Arlo declined to comment for this article, but the company’s terms of service states that it doesn't let its employees access your footage “without your permission unless compelled by law,” or unless you make the footage publicly available.

D-Link and TP-Link didn't respond to CR’s questions, and their privacy policies and terms of services don't provide any details about how they may handle your video footage or who may have access to it.

Honeywell Home and Samsung didn't respond to CR’s questions, but vague language in their terms of service and privacy policies could be interpreted to give them the legal right to watch your camera footage, according to Brookman. In Honeywell Home's privacy policy, for example, the company reserves the right to use audio and video clips "to improve our products and services."

The takeaway from all this? “It can be really hard to nail down what the tech industry is actually doing,” Brookman says. “The good news is some tech companies are being more explicit about their privacy practices after all the recent negative media attention.”