Zoom denies training AI on calls without consent
A backlash over fears that Zoom trained its artificial intelligence (AI) models on customer calls led Zoom to update its terms of service.
It stressed in a blog post that audio, video, and chat data were not used for AI purposes without consent.
A video-calling app acted after users noticed changes to its terms of service in March, which they feared enabled AI training.
To increase transparency, the firm made the changes.
In June, Zoom introduced new AI-powered features, including the ability to summarise meetings without recording them. Trial versions of the features were available for free.
According to some experts, Zoom may have gained access to more user data than necessary because of the terms of service’s original wording.
Robert Bateman, a data protection specialist, said before the terms of service were updated: “The terms seemed to give the service provider a lot of freedom to use data.”
Despite the uncertainty over the risks, he warned that “alarm bells should ring when you encounter broad contractual provisions”.
Zoom updated its terms late on Monday with the line “Zoom will not use audio, video or chat customer content for artificial intelligence training without your consent”.
Artificial intelligence applications are computer programs or tools that perform specific, intelligent tasks that would normally be performed by humans. In order to mimic human-like behavior, they are taught using vast amounts of data and algorithms.
However, the mass extraction of online information for training AI applications has sparked concerns, and prompted lawsuits, over the inclusion of personal, sensitive or copyrighted information in the datasets.
This year, Zoom increased its focus on AI products in response to growing hype about the technology.
According to Open Rights Group, which campaigns on digital privacy, Zoom’s decision to launch the features as a free trial and encourage customers to “opt in” makes the changes “more alarming”.
“While Zoom states that customers will be asked for consent to train AI models on their data, Zoom’s privacy policies are opaque, and it is unclear whether this is the case,” said Abby Burke, policy manager for data protection.
Customer content may be shared with Zoom for “product improvement purposes” upon customer consent, according to a Zoom spokesperson on Monday.
Users joining meetings using AI tools receive warning messages, giving them the option of agreeing or leaving the meeting. Examples of these warning messages were included in Monday’s blog post.
According to Zoom’s chief product officer, Smita Hashim, account owners and administrators could choose whether or not to turn on these features, which are still being trialed, and would “be presented with a transparent consent process for training our AI models using your customer content” if they turned them on.