Zoom Changes TOS to Say It Won't Train AI on Your Calls 'Without Your Consent' After Backlash

After a public outcry, on Monday video conferencing platform Zoom changed its terms of service to try to quell fears that people’s video and audio are being used to train artificial intelligence algorithms. 

On Sunday, people started posting screenshots of a portion of Zoom’s terms of use that states users consent to the company’s use, collection, and storage of “service generated data” for any purpose, including for “training and tuning of algorithms and models.” The TOS defines this data as telemetry, product usage data, diagnostic data, and “similar content.”

Another portion of Zoom’s TOS states that users grant the service the right to use “customer content” (i.e., content generated by users) for the purposes of “machine learning, artificial intelligence, training, testing,” among many other uses. 

It’s a safe bet that most people posting on Twitter have also been on a Zoom call at some point in the last three years, as the pandemic drove work, healthcare, and happy hours to video conferencing beginning in 2020. People were upset, with many saying they would stop using Zoom or were canceling their Pro subscriptions. 

A spokesperson for Zoom told Motherboard that the terms of use in question were updated in March. “Zoom customers decide whether to enable generative AI features, and separately whether to share customer content with Zoom for product improvement purposes,” the spokesperson said, and directed Motherboard to a company blog that was posted on Monday and explains how Zoom uses customer data for AI. 

The blog post explains that Zoom’s generative AI offerings, which include chat composition features and transcription, have an opt-in setting for administrators to choose whether they want data from meetings to be used to “improve the performance and accuracy of these AI services.”  

“We do not use audio, video, or chat content for training our models without customer consent,” Zoom states in the blog post. 

On Monday afternoon, Zoom also added a line to its terms of use to include: 

“Notwithstanding the above, Zoom will not use audio, video or chat Customer Content to train our artificial intelligence models without your consent.” 

While broad-sweeping privacy policies are an unfortunate norm in technology companies today, this is still a striking stance for a company that’s been hit with many privacy criticisms—and a “historic” $85 million lawsuit for Zoombombing—in the past few years. Zoom’s revenue grew 326 percent from 2020 to 2021, but that pandemic-fueled growth has slowed down in recent years, and the company laid off 15 percent of its workforce earlier this year.