Getting on a Zoom call seems commonplace now in a time where remote working is allowed, but the push towards artificial intelligence in certain sectors of tech is making its way into video conferencing app Zoom, and the result may be having the video, audio or chats you’ve inputted into Zoom used to train AI.

Unless, that is, the person who sets up your meetings – and not necessarily you yourself – opts out of AI training in the first place.

Zoom on Monday, August 7, updated its stated terms of service in an attempt to clarify an earlier set of clauses it made which would seemingly allow it to train AI on user data.

While Zoom says it reserves the right to train AI on users’ calls without explicit permission, it added statements to the terms of service promising to seek permission before doing so.

How does this contradiction work? Let’s take a look.

Terms of service adjustments

In a blog post on Zoom, Smita Hashim, the company’s chief product officer, explained how the latest terms of service adjustments outlined the types of content that are created in the process of having a Zoom call.

Gizmodo, in its report, distilled the explanation to its essence, with video, audio, and chat transcripts making up “customer content” and “service generated data” being made up of telemetry data, product usage data, diagnostic data, and additional information gleaned from analyzing content and behavior.

The ‘customer’ versus ‘end-user’ conundrum

Zoom’s terms of service notes that in its customer license grant (under section 10.4 of the terms of service) the customer or “You” is defined as “the person accepting this Agreement as an individual or the legal entity for which the representative is acting” meaning that the customer isn’t always necessarily an end-user for a given Zoom call, but rather is likely to be the one who sets up the call and its permissions.

The latest update to Zoom’s terms of service has this particular clause added now as part of its customer license grant: “Notwithstanding the above, Zoom will not use audio, video or chat Customer Content to train our artificial intelligence models without your consent.”

Gizmodo also added background information on a set of features that can be instituted in a call, namely Zoom IQ, which can summarize chat threads and help generate automated responses to written chat questions during a call.

Gizmodo adds, “When you enable these features, there’s a little check box that’s turned on by default. If you don’t bother to change it, you agree to let the company collect data to build and improve its AI. When a call starts with Zoom IQ enabled, other people in the call get a notification about it titled ‘Meeting Summary has been enabled.’ The popup says “The account owner may allow Zoom to access and use your inputs and AI-generated content for the purpose of providing the feature and for Zoom IQ product improvement, including model training.'”

This means that users need to leave a call if they do not consent to being used for AI training, as the creator of an AI-enhanced Zoom call – the administrator of the call or the “customer” – has already consented on your behalf as an end-user engaging in call, unless you decide not to join in.

Meanwhile, Zoom has not responded to queries regarding opting out of service generated data being used to train AI. There also does not currently appear to be a way to opt out of having the service generated data used for training AI.

Zoom backtracks

On August 11, Zoom walked back its updates regarding AI training on its terms of service following online feedback. 

The company’s original blog post on the matter has since been updated to state that “Zoom does not use any of your audio, video, chat, screen-sharing, attachments, or other communications like customer content (such as poll results, whiteboard, and reactions) to train Zoom’s or third-party artificial intelligence models.” – Rappler.com