Several posts on Facebook have claimed that the video conferencing software Zoom’s terms of service “now require you to allow AI to train on ALL your data—audio, facial recognition, private conversations—unconditionally and irrevocably, with no opt out”.
This is not what Zoom’s current terms of service say, or what they said at the time these claims were made on Facebook, although its terms of service did previously appear to suggest service-generated data and content could be used by AI. The terms now say clearly that it 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 its own or any third-party AI. Zoom also told Full Fact it didn’t use customer audio, video or chat to train AI in the period when its previous terms appeared to imply it could.
False claims about online privacy may cause unnecessary alarm. We’ve previously checked similar claims that all Instagram posts and messages were being made public and that the UK Emergency Alerts were breaching GDPR.
Double your donation
Give today via The Big Give and your donation will be doubled.
Double my donation
Where did the claims come from?
Claims about Zoom using meeting footage to train AI stem from changes made to Zoom’s terms of service first noted in a 6 August tech blog post.
A post on X (formerly Twitter) which has been viewed two million times and links to that blog post says: “Well time to retire @Zoom, who is [sic] basically wants to use/abuse you to train their AI.”
Zoom’s terms of service were changed at the end of March. An archived version of the terms from 31 March makes no mention of AI or artificial intelligence, but from 2 April onwards there were two references to it.
It said: “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)”.
It also said: “You agree to grant and hereby grant Zoom a perpetual, worldwide, non-exclusive, royalty-free, sublicensable, and transferable license and all other rights required or necessary to redistribute, publish, import, access, use, store, transmit, review, disclose, preserve, extract, modify, reproduce, share, use, display, copy, distribute, translate, transcribe, create derivative works, and process Customer Content and to perform all acts with respect to the Customer Content [...] for the purpose of product and service development, marketing, analytics, quality assurance, machine learning, artificial intelligence, training, testing, improvement of the Services, Software, or Zoom’s other products, services, and software, or any combination thereof”.
Terms of service updated
After the tech blog post pointing out this change was published, at some point on 7 August or 8 August Zoom’s terms were updated to add a line saying: “Notwithstanding the above, Zoom will not use audio, video or chat Customer Content to train our artificial intelligence models without your consent.”
Zoom also published a blog post on 7 August clarifying what it meant, saying: “For AI, we do not use audio, video, or chat content for training our models without customer consent.”
Then on 11 August the terms of service were changed again to remove the three previous original references to artificial intelligence, and add the line: “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.”
At the time of writing the terms still say this.
What Zoom has said
Zoom’s own blog post was also updated to reflect the new changes. It said: “Following feedback received regarding Zoom’s recently updated terms of service Zoom has updated our terms of service and the below blog post to make it clear 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.”
On 8 August, Zoom’s CEO Eric S. Yuan also posted on Linkedin to “set the record straight”.
He wrote: “If you use our two new generative AI features that have been available on free trial for two months, you can see in the UI [user interface] that we definitely ask customers if they want to opt-in to share content to improve our products and train our AI models.”
Zoom says these two AI features, one that summarises meetings and one that drafts messages using context from the conversation, need users to enable them and can be turned off.
Zoom confirmed to Full Fact that these features don’t use meeting content to train AI models, instead using data Zoom has purchased or created, or which is in the public domain.
On why the terms of service were changed in March, Mr Yuan said: “We had a process failure internally that we will fix. Period.”
A spokesperson for Zoom told Full Fact that, even before the most recent update to its terms of service on 11 August, “Zoom did not use customer audio, video, chat, screen-sharing, attachments, or other communications like customer content (such as poll results, whiteboard, and reactions) to train our AI models”.
Image courtesy of Chris Montgomery