A Closer Legal Look: Zoom’s Pivot to AI Model Training with User Data
Introduction
In the 21st century, Artificial Intelligence (AI) has become commonplace. Some view AI as a valuable aid in enhancing and streamlining daily tasks. However, others see it as a crutch for the lazy and reliant. As with any coin, AI has both positive and negative aspects.
All the major Tech giants, be it the ChatGPT created by OpenAI or Google Bard, a generative AI chatbot created by Google that uses a similar machine learning process and AI tool that reduces human effort. In September, Zoom announced the launch of its new generative AI tool, ‘Zoom AI Companion,’ which uses its algorithm to generate real-time feedback in Zoom video conferences/meetings to improve efficiency.
Zoom’s new AI Companion: what is it?
The ‘Zoom AI Companion’ is a generative AI assistant that assists the user in real-time by generating feedback in terms of easing out tasks that require human intelligence to struggle in a short time frame, provides an uninterrupted meeting platform, assisting all those late-comers to catch up with a small brief without interrupting the meeting, and much more through AI Companion’s use of its machine learning technique through transmitting relevant information more simply for the users to experience smooth-flow of their work.
Some of the things that Zoom AI Companion can do:
- Compose messages/emails just like any other AI according to the need.
- Generating summaries of meetings/messages to improve efficiency and ensure the users are not lagging in their meetings/ workplace.
- Real-time feedback in live meetings, such as summary for late joiners to stay on the road, smart recordings
- Categorization of data/ generating valuable ideas.
How are Zoom’s new AI features raising privacy concerns?
Zoom has been in controversy with allegations involving how the video conferencing platform uses the personal data of its users to train its AI models, which had led to the change in its Terms of Service (ToS), where Clause 10.2 provides that Zoom does not use the data of its users and at the same time provides
“Permitted Uses” reads as:
“10.2 Permitted Uses and Customer License Grant. Zoom will only access, process, or use Customer Content for the following reasons (the “Permitted Uses” )::
- Consistent with this Agreement and as required to perform our obligations and provide the Services.
- By our Privacy Statement.
- As authorized or instructed by you.
- As required by Law.
- For legal, safety or security purposes, including enforcing our Acceptable Use Guidelines.
You grant Zoom a perpetual, worldwide, non-exclusive, royalty-free, sublicensable, and transferable license and all other rights required or necessary for the Permitted Uses.
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.”
These terms state that two types of data are being generated using the Zoom platform: “Customer Content” and “Service Generated Data”. The restrictions mentioned in Clause 10.2 only apply to Customer Content, which includes audio, video, chat, screen sharing, and attachments, which can also be used subject to users’ consent, so it is clear that the user’s data is stored and constantly monitored and Zoom has unrestricted access to it. However, the platform does not specify any limitations on using Service-generated data. The AI model is permitted to use this data without any restrictions.
According to Clause 10.5 of its Terms of Service, Zoom collects Service Generated Data from its users globally, which includes Telemetry data, product usage data, diagnostic data, and other similar data. Zoom claims exclusive rights over this data and can use it in any way without any restrictions.
Zoom had updated ToS only after the anti-competitive practices allegations against it were made that the platform trained its AI by using sensitive data of its users; however, after much resilience, Zoom responded with a statement clarifying that the platform does not use any of the data like audio, video, visuals to train its own AI or any similar third-party AI.
In its updated terms, the company has assured that the platform is even more transparent and open to its users regarding how their data is not used by its AI or any third-party AI. Little did they know that the platform tracks their behaviour and usage patterns (Service Generated Data) without their consent and poses serious threats to the privacy of its users. The Customer Content collected from users is only used with their consent to enhance AI capabilities, while service-generated data is unregulated.
Zoom’s previous record in privacy concerns
There have been concerns about Zoom’s privacy policy in the past and allegations similar to those of Open AI’s tool ChatGPT, which had caused severe regulatory breaches in terms of invasion of privacy laws in many countries by using personal information from millions of people, creating privacy tensions around these AI tools. Italy was the first country to impose sanctions against Open AI’s ChatGPT and prohibited the AI tool from using its citizens’ personal information.
Zoom was involved in a similar scandal where multiple allegations of misleading the users and the platform’s data security policies were questioned. The Federal Trade Commission (FTC) also investigated Zoom and found zero transparency in its data usage policies.
Zoom was under the FTC radar for a significant period, and the FTC had severe concerns over its data usage practices and whether the company was involved in any unfair trade practices. A settlement between the two arrived where the video conferencing platform had assured that they would release updated terms (as stated above) being more clear and transparent in terms of its data usage policies. However, the updated policies only restrict Zoom from using customer content and do not regulate the usage of other data types.
Conclusion
Although AI generative tools are a creation of human intelligence, concerns remain about the privacy of users who utilize them. There is still a debate about why AI tools are so popular in the 21st century. Furthermore, it remains to be seen if these tools will be regulated. Similar to Zoom, where data collection and learning is only done with user consent, it is also flawed, where the platform stores multiple types of data as per Clause 10 of its ToS. The development of AI is an ongoing process, and while AI tools are just the beginning of a new era of development, it’s still unclear what the future of AI holds.
Frequently asked questions
Can Zoom use the audio and video from its calls to train its AI?
No, Zoom cannot use the audio and video from its calls to train its AI because Clause 10.2 of its Terms of Service provides that Zoom does not use the audio and video to train Zoom or third-party artificial intelligence models.
Is Zoom using your meetings to train its AI?
Zoom is restricted, as per Clause 10.2 of its updated Terms of Service, to train its AI or any third-party artificial intelligence models via using the content of the meeting of its users.
However, suppose the meeting data falls within the ambit of Service Generated Data. In that case, the content of the Meetings is used without users’ consent to train its AI and for other purposes best known to the platform only.
