Who's Zoomin' Who?

Zoom wants to use your data to train its AI. Is that a bad thing? Maybe. [UPDATE 8/10: Zoom is changing its policies to shield data from its AI. You're welcome.]

When robots Zoom. Source: DALL-E.

Like a lot of people, I became a little addicted to Zoom during the pandemic. [1] For a while there it felt like my only porthole onto the larger world. And even after the lock down lifted I continued to use it. I've moved away almost entirely from doing phone interviews; I do them all face to face, record them, and use an AI-based service to transcribe the conversations. I also still use it for out-of-office medical visits and occasional catch ups with distant friends.

(Zoom is also sometimes unintentionally revelatory, in a "Jeez I really need a haircut" or "When did my neck get so wobbly?" kind of way. [2])

Over the past weekend Zoom made a few changes to its Terms of Service (TOS) that a few bloggers noticed and set everyone's hair on fire. Specifically, Zoom amended its TOS to allow it to use some of the data it collects to train AI models. The changes were made to accommodate a few new generative AI-based add-ons Zoom has introduced that can automatically summarize meetings or compose chat messages.

But the immediate assumption by many outside observers was that Zoom was reserving the right to use the content of Zoom conversations to train its AI. And given how many confidential conversations happen on Zoom, that caused some folks (like the extremely accomplished investigative journalism organization Bellingcat) to immediately cancel their paid Zoom subscriptions.

Source: NBCnews.com

Because as we've recently learned, generative AI models are terrible gossips. They slurp up confidential or proprietary information and then spill it to anyone who asks. And they do it under the guise of using this data to continue to train their models to be better at answering questions. This is why 75 percent of global organizations are considering banning the use of these tools at work, according to a new survey by BlackBerry. [3] They want to keep ChatGPT from inadvertently spilling their tea.

Technically, that's not what Zoom is actually doing. At least, not yet. But what they are doing shouldn't make you feel warm and fuzzy either.

Terms yes; endearing, no

Zoom's new TOS does two things. The first is that it reserves the right to use "service generated data" to train its AI. This is not the content of conversations, but more of the metadata around them — when they happened, how long they lasted, whether there were technical glitches, etc. Zoom says this doesn't include any personal information, and they can do whatever the hell they want with it. Fair enough.

The second thing the TOS says is that when you use Zoom you grant it a "perpetual, worldwide, non-exclusive, royalty-free, sublicensable, and transferable license” for various purposes, including “machine learning, artificial intelligence, training, testing, improvement of the Services, Software, or Zoom’s other products, services, and software, or any combination thereof.”

It then adds (in boldface, so you know they're serious):

Notwithstanding the above, Zoom will not use audio, video or chat Customer Content to train our artificial intelligence models without your consent.

That "without your consent" bit is the part that alarms me. It suggests that at some point, they will ask for that consent. But how, exactly? Nowhere in Zoom's TOS, Privacy Policy, or the blog post company Chief Product Office Smita Hashim felt compelled to publish yesterday does it say how Zoom plans to obtain its users' consent, whether users are opted-in by default, or what services may or may not be available if they decline.

Privacy & security wonk Richard Henderson is asking the same questions:

Explicit opt-in? Explicit opt-out? Implied consent when you accept their Terms of Service? What sort of consent is Zoom talking about here? What about guests in a meeting where the only "choice" is to leave the meeting? That's not much of a choice.... Imagine this scenario: you join a meeting, and someone turns on the AI features and your only option is to click “ok” or “leave”… people are going to feel socially coerced to accept it even if they feel uncomfortable. Who wants to stand out like that? Not many.

I've sent questions to Hashim (via LinkedIn) and to Privacy@Zoom.us. If I get a response, I'll update this post. But I'm not holding my breath waiting for it.

When Zoom goes boom

My gut tells me that this is less evil than it sounds, and it's a really just another episode of Lawyers Gone Wild. Someone in Zoom's legal department is trying hard to carve out wiggle room for the company by reserving as many rights as possible, [4] in case Zoom does decide at some point it wants to spoon feed more customer data to the AI creature living under the bed. Still, the company's response to this kerfuffle has been more than a little underwhelming.

While researching this post, I decided to ask Zoom for a copy of the data it had collected about me. To get there, I had to jump through more hoops than Shamu the Killer Whale. Then the form I had to fill out refused to work at first, and when it finally did, the Zoom Privacy Portal told me that "No text was entered into the Request Details field by the individual." I sent a message to Zoom support asking why. Still waiting on a response to that, too.

Finally, I looked at a third-party app found in Zoom's App Marketplace, Read.ai. That app appears to do the same thing that Zoom's Meeting Summary tool does. I thought maybe I'd see some kind of opt-in form. Instead, I saw this:

So... If Zoom doesn't trust the websites of the apps in its own marketplace, why on earth would any of the rest of us trust Zoom?

UPDATE: After a lengthy discussion about this on LinkedIn, Zoom CEO Eric S. Yuan announced that the company WILL NOT use the content of any conversations for training any of its AI:

After considering all of your feedback, we are committing to all of our customers that we will never use any of their audio, video, chat, screen sharing, attachments and other communications like poll results, whiteboard and reactions to train our AI models.

Did this happen shortly after I suggested they do this? Yes. Am I taking credit for it? I am far too modest and humble for such things.

Do you still Zoom the way you did last summer? And will you continue to do it now? Share your thoughts in the comments below; they will not be used to train anything, except possibly my cat.

[1] Also to day drinking and excessive hand sanitizer use. I have like a gallon of the stuff now, just sitting in my bathroom cabinet, if anyone needs any.

[2] I recommend turning off 'self view.' Does wonders for one's self esteem.

[3] Yes, that BlackBerry. They're still around. Who knew?

[4] In my experience, if you ask a lawyer how to change a light bulb, they'll make you sign a 20-page liability waiver absolving them of any responsibility if you get electrocuted. And then they'll send you a bill.