Zoom skyrocketed in popularity during the early days of the COVID-19 pandemic, as millions of people around the world suddenly needed to work from home and have meetings over video calls. However, a recent change to the company’s terms of service sparked concerns about training AI with user data.
Zoom’s latest terms of service agreement, which went into effect on July 27, 2023, outlines that Zoom may compile “Service Generated Data” based on the customer content and how the software is used. Section 10.2 of the agreement specifically states that the generated data may be used for machine learning or artificial intelligence:
You consent to Zoom’s access, use, collection, creation, modification, distribution, processing, sharing, maintenance, and storage of Service Generated Data for any purpose, to the extent and in the manner permitted under applicable Law, including for the purpose of product and service development, marketing, analytics, quality assurance, machine learning or artificial intelligence (including for the purposes of training and tuning of algorithms and models), training, testing, improvement of the Services, Software, or Zoom’s other products, services, and software, or any combination thereof, and as otherwise provided in this Agreement.
The new agreement is alarming because many other companies have been training generative AI with user data and content without direct authorization from the user. Google confirmed in July that it’s training Bard and other AI products using scraped web data, and tools like Stable Diffusion and ChatGPT also rely on data that wasn’t authorized for use in AI products.
After enough people pointed out the alarming agreement, Zoom published a blog post explaining its use of AI. The company claims, “for AI, we do not use audio, video, or chat content for training our models without customer consent.” However, that explicit language doesn’t appear to be in the terms of service, so that policy could theoretically change without an update to the terms of service.
Zoom has two generative AI features right now: Zoom IQ Meeting Summary, and Zoom IQ Team Chat Compose. They add meeting summaries and AI-powered chat composition, but (at least for now) they are turned off by default. When you turn them on, there is an additional switch to enable data sharing for training AI models. If you choose to share data, that data is only used for internal AI training and is not used for training third-party models, according to Zoom. There are also popups in calls where Meeting Summary is enabled, which tells all participants that the call will be processed with AI.
For the moment, it doesn’t seem like Zoom is using AI in an outright harmful way, but it might be worth looking into alternatives anyway — the service has had other problems over the years. Jitsi is a popular open-source alternative, which can even be hosted on your own server if desired.