Our favourite pandemic-era technology Zoom has had to walk back some significant changes to its terms of service, but the artificial intelligence wars have just begun.
Zoom was catapulted from relative obscurity to mainstream use during the pandemic to provide convenient and user-friendly videoconferencing, for socialising, education or professional purposes. The rise of Zoom was so powerful that it entered our language: “let’s zoom them”. This is all the more remarkable as unlike most major technology platforms, it is not owned by any of the so-called MAAMAs: Meta (Facebook), Alphabet (Google), Amazon, Microsoft and Apple.
Although it has no familial ties to these tech titans, it is also no stranger to privacy and data-related controversies [https://www.vox.com/recode/2020/3/31/21201019/zoom-coronavirus-privacy-hacks]. In 2020, the U.S. Federal Trade Commission launched an investigation into the company’s use of keys to access the content of user meetings. Zoom promised to would implement security programs and tests as part of the settlement [https://www.ftc.gov/system/files/documents/cases/1923167zoomacco2.pdf].
Now it has landed in hot water again.
In late July, changes [http://web.archive.org/web/20230804010858/https:/explore.zoom.us/en/terms/] to the Zoom terms of service came into effect. To understand these changes, it is important to know that Zoom collects two types of data.
The first are “Service Generated Data”, which relate to customer usage data such as diagnostics, user location, and software interactions. Zoom now required customers to provide consent for it to use, distribute and store this data. The second type of data are “Customer Content”, which includes uploaded files, meeting transcripts, visual displays, or polls. Zoom now required customers to grant it a very wide licence to use, transmit, modify or store that data.
Consent or licence to do what? Unsurprisingly, a huge range of things. But one of those things was “product and service development”, including training of machine learning or AI. Consumers soon caught wind of the changes and their fury combusted.
This techlash led Zoom’s Chief Product Officer, Smita Hashim, to rapidly engage damage control by blog post. On 7 August, Hashim clarified [http://web.archive.org/web/20230807115434/https://blog.zoom.us/zooms-term-service-ai/] that the change about service generated data was to deliver better “user experience”, such as by managing server loads according to time zones. The change about customer content was to add new services, such as meeting records, and to enhance service safety, such as by automated scanning for spam and fraud. She also introduced two new opt-in generative AI features, Zoom IQ Meeting Summary and Zoom IQ Team Chat Compose. These tools enable “automated meeting summaries and AI-powered chat composition”.
Of course, consumers should pay attention to what services can do, not what they say they will do. Hashim may talk about better services, but the terms clearly did not care about customer consent when using data. This may be why the blog post was substantially edited [https://blog.zoom.us/zooms-term-service-ai/].
The editor added a further note:
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.
Zoom then changed its terms of service again. The new terms [https://explore.zoom.us/en/terms/] clarify that customer content can only be used for delivering the services, unless authorised by customers. However, Zoom “owns all rights, title, and interest in” the service generated data. It learned that sometimes, less detail is more.
Zoom’s CEO Eric Yuan admitted “we had a process failure internally”. A process failure over four months that was only identified when it made headlines.
These events involved one company, but others such as Google [https://www.theverge.com/2023/7/5/23784257/google-ai-bard-privacy-policy-train-web-scraping] or the web browser Brave [https://stackdiary.com/brave-selling-copyrighted-data-for-ai-training/] are doing similar things. Like many companies, Zoom is trying to walk a tightrope over disclaiming responsibility for the content that is transmitted on its services but then claiming rights to use it.
The episode illustrates three broader truths about the precipice we stand on right now.
Firstly, the Zoom story is a glimpse into what our world might look like this century. As intellectual property lawyers, such terms are standard in the business-to-business contracts we see. As part of the Fourth Industrial Revolution, our online interactions are now monitored and increasingly being used to train AI models, thanks to stealthy terms of service. Companies all the time want to use customer data generated to make reports that they sell to back to consumers. Companies are increasingly introducing these updates to terms of services by stealth.
For Zoom, that means your calls to your grandmother overseas, your doctor’s appointment, your legal advice, or the class you painstakingly prepared for your university course. If our interactions now are mediated by large corporations, our ownership of those interactions and what they create will be a big battle in the 21st century. Fortunately, there are reforms governments can make [https://papers.ssrn.com/sol3/papers.cfm?abstract_id=3448959] to intellectual property law to ensure consumers have a greater say over these uses, and they are properly compensated for it.
Secondly, the episode shows people’s current discomfort with the AI contract. By and large, we willingly hand over data to Google Maps, YouTube or TikTok if it means we get entertainment or convenience back. But this time consumers come to new technologies forewarned and highly sceptical. The technology’s benefits have not become necessities, and there are a lot negatives weighing on people’s minds [https://www.bbc.com/worklife/article/20230418-ai-anxiety-artificial-intelligence-replace-jobs]. The public relations job for tech companies is much harder for these technologies than it was, at least for now.
Lastly, it shows that consumers still hold power. Zoom is an interesting tech giant, because unlike say Microsoft it does not many other products. It has many competitors. Zoom knows that public perceptions can change very quickly. It had a moment of hubris, but its fear of losing consumers to the gravitational pull of network effects for its competitors led to it make a hasty retreat. When consumers, consumer watchdogs and news media coordinate, the power of the tech titans is not inevitable.