Like everyone else, Zoom has added AI features to improve its app and video conferencing service. We all love the ease and speed AI provides, but there are often concerns about the data used to train models. Zoom is training its AI on customer data. Here’s what you need to know.
News leaked in May 2022 that Zoom was working on emotion-sensing AI that could analyze faces in meetings. Beyond the potential for inaccurate analysis, the results could be discriminatory.
More recently, privacy concerns have arisen since Zoom updated its terms of service to allow the use of your data to train its AI model. Following a backlash, Zoom adjusted the language to provide more transparency about this issue and clarified how the option works.
In bold text, the update reads, “For AI, we do not use audio, video, or chat content for training our models without customer consent.” It’s that last part about consent that’s a little tricky. Zoom won’t use your data to train its models across the board, but it will if you consent. That consent is automatically applied when you engage a Zoom IQ feature. You can still use Zoom IQ without your data being harvested for training, but you have to manually opt out once the AI features are enabled.
The Zoom IQ AI feature is off by default. Meeting administrators can turn it on to generate a summary when the meeting ends. When enabled, anyone joining the call will see a pop-up notification alerting them, “Meeting Summary has been enabled.”
You can enter the meeting, implicitly giving consent, or leave the meeting. That might be a difficult choice if it’s a work meeting. Can you afford to skip a meeting? That’s an issue to raise with the meeting administrator.
Sections 10.2 and 10.4 of Zoom’s terms of service grant full rights to store and distribute your data for training and tuning its AI models. A notwithstanding clause appears after 10.4 to clarify that “Zoom will not use audio, video or chat Customer Content to train our artificial intelligence models without your consent.”
There’s a legitimate concern about sharing too much data with an AI. It’s relatively easy to create a deepfake of your voice or face with AI.
Digital Trends reached out to ask Zoom what consent looked like for training its AI model, and it provided the following statement:
“Zoom generative AI features are default off. Zoom account owners and administrators control whether to enable these AI features for their accounts and can opt out of providing their Customer Content to Zoom for model training at the time of enablement. Customers can change the account’s data sharing selection at any time.”