Zoom Promises It Won't Use Your Videos And Chats To Train AI

After its updated terms of service started raising eyebrows, Zoom has revised its wording and clarified in a separate blog post that it will not use customers' data to train AI. That data includes the faces and voices of customers and other chat participants using its extremely popular video calling and conferencing software. 

In the company's listed terms of service, updates were made to Section 10: Data Usage, Licenses, and Responsibilities, and includes this added language: "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 or third-party artificial intelligence models."

The clarification was added after the prior use of the phrase "without your consent" in Zoom's terms of service, which was revised last March. This struck alarm bells with many people because it implied that, in some situations, Zoom could be using customers' "communications-like" data to train AI. Even worse, how that consent could be given or revoked wasn't laid out. 

In addition to updating Section 10, Zoom Chief Product Officer Smita Hashim elaborated on the changes in a company blog post to further assuage customers' concerns. The blog post both reiterates that customers' data won't be used to train any AI models, and that the company will "strive to provide transparency about data ownership in our terms of service."

Currently, Zoom uses generative AI for certain tools available to customers like Zoom IQ Team Chat Compose and Zoom IQ Meeting Summary. The latter can monitor the video, audio, and text chats of a Zoom conference to then provide a summary of the meeting. 

Zoom isn't the only company facing scrutiny over what it uses to train its AI. Whether it's businesses looking to use AI in specific use-cases, or companies like OpenAI that are building generative AI platforms like ChatGPT, the sources of the content used to train artificial intelligence are being looked at closer than ever.

Concerns over AI training data have been focused on privacy, such as Zoom users worried their faces and conversations are being used, as well as copyright issues. Comedian Sarah Silverman — whose memoir, "The Bedwetter," was allegedly used without her consent by OpenAI to help train ChatGPT — has joined other authors in a copyright lawsuit against the tech company. 

Other similar protests are growing as the public learns more about how generative AI models are built. Even the Screen Actors Guild has raised concerns over actors' voices and likenesses potentially being digitally created by AI without their consent.

Looking to avoid any negative press, Zoom went out of its way to diminish users' uneasiness over its own AI practices, and other companies will likely do the same. However, as generative AI advances and the technology is used in more industries, it's likely that more debates and legal fights over AI content will continue to proliferate.