Zoom can now use your data to train its AI, even though it says it won’t. Here’s what you need to know.

A terms-of-service update from Zoom — the popular videoconferencing software — allows the company to use customer data to train its artificial intelligence. The update is giving some users pause, prompting the company to say it won’t actually use the data without permission.

As reported on by Stack Diary, a Zoom terms update in March said it reserved the right to train its AI on user data. There’s no clear way to opt out.

As the update continued to roll out, more people began speaking out about privacy concerns — especially about the use of the software in settings like therapy and legal matters.

But in a blog post released by Zoom on Monday, the company said it isn’t actually training its AI on customers’ video calls, despite its new service terms saying it can.

“Our intention was to make sure that if we provided value-added services (such as a meeting recording), we would have the ability to do so without questions of usage of rights,” the company blog said. “For AI, we do not use audio, video, or chat content for training our models without customer consent.”

It’s worth noting that many video and audio platforms — including Google Meet and Teams Premium — already rely on AI technology to provide live closed captioning services.

Still, Zoom’s blog post attempting to clear up concerns instead appears to have sparked new attention to the term updates.

On social media, users speculated that large companies wouldn’t want to risk potential leaks of internal content.

TV and film writer John Rogers suggested that it could lead to HIPAA issues in the medical field (while some health-care providers use Zoom, many already use internal platforms or other alternates specifically designed for the setting) as well as TV and movie studios dropping the service.

The Writers Guild of America and the actors’ guild, SAG-AFTRA, are both on strike, bringing Hollywood essentially to a halt — because of contract negotiation issues, including concerns with the use of AI.

The updated terms — presented to Zoom users in fine print — became effective July 27.

They give Zoom the right to use some elements of customer data to train its AI, a fairly common move for tech companies these days, according to CNBC.

The terms say customers 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 ... machine learning or artificial intelligence (including for the purposes of training and tuning of algorithms and models).”

Still, the company reiterated in its blog Monday, it won’t use audio, video, or chat content to train its models without customer consent.

“We remain committed to transparency, and our aim is to provide you with the tools you need to make informed decisions about your Zoom account,” the blog post said. “We value your privacy and are continuously working to enhance our services while respecting your rights and preferences.”

As noted by CNBC, the discourse about Zoom’s update comes amid a growing debate regarding how AI services should be trained and the ethics surrounding it all.

Artists have described concerns with AI graphic generators and how the technology can use existing work to train itself along with its potential to replace human labor.

» READ MORE: Lensa is the app behind your friends’ new profile pics. Here’s what artists have to say about it.

“It’s a hard battle because it seems fun and harmless, but I think this could be, and in a lot of ways already is, the beginning of some serious problems in all creative fields,” said Philly-based graphic designer and illustrator Nicole Saltzer.

And it’s not just image-generating tools that are raising eyebrows. Chatbots, like ChatGPT and Bard by Google, are trained on internet text, leading to scrutiny of/ what’s up for grabs and who is at risk.

In June, Zoom brought on two new and free AI features. One summarizes meetings and the other composes chat messages. A user has to enable the features. If they opt-in, they are prompted to sign a consent form that allows Zoom to train its AI models using their customer content (which can include your video, audio, and chat transcripts).

“Your content is used solely to improve the performance and accuracy of these AI services,” Zoom’s blog post said.

A company spokesperson told Gizmodo that the power is in Zoom users’ hands.

“Zoom customers decide whether to enable generative AI features,” they said, referring to the two new AI features, “and separately whether to share customer content with Zoom for product improvement purposes.” The company declined to answer Gizmodo’s questions about its use of “Service Generated Content,” which includes analytics data and anything besides the already-outlined customer content.

Gizmodo also noted that Zoom’s track record on how it honors promises about privacy is shaky.

In 2020, the company said it would only offer end-to-end encryption to paying customers. A lawsuit claimed that Zoom said it offered that type of privacy already to all users. The platform ended up fixing the discrepancy and strengthening its encryption for basic users. A year later, the company agreed to an $85 settlement after sharing data with Google and Facebook without telling customers.