Zoom Insists It Wasn’t Using Private Calls to Train AI System

Nothing to see here, folks!

Zoom With a View

Zoom wants you to know that it definitely, 100 percent hasn't been using your video calls to train its AI — even if its convoluted Terms of Service agreement seems to strongly suggest otherwise.

To back up a bit: the pandemic-era video conferencing service came under fire this week on grounds that, during its March launch of its generative AI-powered "Zoom IQ" features, it had rewritten its TOS to essentially make any and all "Customer Content" open-season for AI training.

By checking the agreement box, the updated TOS reads, users "consent to Zoom's access, use, collection, creation, modification, distribution, processing, sharing, maintenance, and storage" of user data for "any purpose" including "machine learning or artificial intelligence (including for the purposes of training and tuning of algorithms and models)."

You can't really get any clearer than that, and Zoom users, who seemed generally unaware of the update until Stack Diary first reported about the changes on Sunday, were unsurprisingly upset by the revelation. After all, if the TOS was updated back in March, how much of their private data — which may have included the content of Zoom therapy or telehealth meetings, corporate meetings, and intimate conversations — had been guzzled up by Zoom's AI?

Backtrack City

In response to user backlash, the video conferencing service updated its policy with an even more confusing promise: "Notwithstanding the above, Zoom will not use audio, video, or chat Customer Content to train our artificial intelligence models without your consent."

Elsewhere, the company also published a blog post about AI transparency, where it further emphasized that it won't vacuum up video call data without user permission. But the "notwithstanding" note in the freshly updated TOS is still pretty vague about what the scope of "Customer Content" allows or disallows.

"By its terms, it's not immediately clear to me what is included or excluded," Chris Hart, the co-chair of the privacy and data security practice at the law firm Foley Hoag, told Vox. "For example, if a video call is not included in Customer Content that will be used for AI training, is the derivative transcript still fair game? The whiteboard used during the meeting? The polls? Documents uploaded and shared with a team?"

Overall, it's been a terrible look for the company, especially considering its past privacy scandals. But everyone wants a piece of the AI gold rush pie, and human-generated data is becoming an increasingly valuable commodity. As we enter this new, AI-laden technological era, maybe let this be a reminder to start checking the TOS.

More on AI and Privacy: Sarah Silverman Sues OpenAI for Copyright Infringement