Our software vendor wants to use our data for AI. What does that mean for us?

Usually that its terms let it improve its own AI features with your data, in some products until an admin opts out. Licensing records to an AI buyer is a separate deal you negotiate and are paid for, even when the vendor is the buyer.

What do our vendors’ terms already allow?

Some let the vendor use your content, or data about how you use the product, to improve its own AI; others rule out training models on your content. Each vendor says which in its own terms or AI privacy page. As of October 2026:

Vendor	What its own pages say	Opt-out or setting
Atlassian (Jira, Confluence)	Since August 17, 2026, uses “metadata” from every cloud plan and, by default on Free and Standard, “in-app data” such as page and work item content; all of it is “de-identified and aggregated before it is used”	Organization admins can turn in-app data off on any plan; only Enterprise can opt out of metadata (Atlassian)
Google Workspace	“Workspace does not use customer data for training models without customer’s prior permission or instruction” (updated August 14, 2026)	Nothing to switch off for training; check whether anyone gave permission
Microsoft 365 Copilot	“Prompts, responses, and data accessed through Microsoft Graph aren’t used to train foundation LLMs”	Nothing to switch off for foundation-model training
ServiceTitan	May use your data to “train, tune, and improve AI models solely to provide the Service to Customer”; its data preferences for product development and benchmarks start switched on (ServiceTitan)	An administrator, under Settings > Titan Intelligence; the ServiceTitan page has the detail
Slack	Analyzes “Customer Data (e.g. messages, content, and files)” to build predictive features such as emoji and channel recommendations, and “will not use Customer Data to train generative AI models unless Customer provides affirmative opt-in consent”	An Org or Workspace Owner or the Primary Owner emails feedback@slack.com with the subject “Slack Global model opt-out request”
Zoom	“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” (terms section 10.2, effective August 11, 2023)	Nothing to switch off for communications content

Where there is a switch, it sits with whoever administers the account. If that is your IT firm, Our IT is outsourced covers who holds which role.

Can a vendor change its terms?

Yes, and three dated episodes show how it happens:

• Zoom, 2023. Social media posts drew attention to a clause in which users consented to Zoom’s use of “Service Generated Data” for “any purpose”, including “machine learning or artificial intelligence”, 9to5Mac reported on August 7, 2023. Zoom’s chief product officer answered in a blog post the same day (Zoom updated it on February 7, 2024), and the terms effective August 11 carry the sentence quoted above.
• Slack, 2024. In May 2024 a Hacker News post linked Slack’s privacy principles, which described training “global models” on customer data. TechCrunch reported on May 17 that opting out meant emailing the company, and that a Slack engineer acknowledged the page needed updating. As of October 2026 the opt-out is still an email.
• Atlassian, 2026. gHacks reported on April 19, 2026 that Atlassian would collect Jira and Confluence data by default to train AI models from August 17. Atlassian’s page says the change took effect that day, and its stated reason begins “By learning from richer, more diverse customer data and usage patterns, we can deliver enhanced AI capabilities”.

Read the current version, not the one someone accepted at sign-up, and note the date you read it.

Is a vendor using our data the same as us licensing it?

No. They rest on different documents and work differently:

	A vendor using your data under its terms	You licensing your records
The document	Terms accepted to use the software. Zoom’s grant, for example, is a “perpetual, worldwide, non-exclusive, royalty-free, sublicensable, and transferable license” for its permitted uses	A license you negotiate and sign
Who is paid	Nobody pays you; you pay for the software	You are
What is used	Whatever the terms cover	The systems, years and record types in the agreed scope
Who can change it	The vendor, by updating its terms	Both sides, in writing
Where it goes	The vendor’s own products; Zoom rules out training third parties’ models on communications content, and ServiceTitan rules out training for third parties	A named buyer, for the uses the license states

The two can sit side by side. Zoom’s grant is non-exclusive, for example; check whether each vendor’s is. The vendor’s rights matter more when a buyer wants exclusivity, which some programs start from (exclusive or not), and because a license’s warranties cover your rights in the records. Give counsel your vendors’ terms along with the draft license.

Which data in our account belongs to the vendor?

Your content usually stays yours; data the software generates may not. Under Zoom’s terms, “You retain all ownership rights in your Customer Content”, while “Zoom owns all rights, title, and interest in and to Service Generated Data”, meaning telemetry, usage and diagnostic data. ServiceTitan’s terms give it ownership of what they call Aggregate Data (ServiceTitan page).

The question has already come up in a sale. In the Spirit Airlines bankruptcy, Springshot, whose aviation operations software Spirit used, asked the court to keep records its software generated out of the data being sold. It cited contract language giving it rights to “any data or information generated by the Service or Software,” and its CEO, Doug Kreuzkamp, separated “the information Spirit supplied” from “the proprietary outputs our platform generated” (The Next Web, September 18, 2026).

Before a system’s records go into scope, read its terms for who owns generated, usage and aggregate data, and leave out anything the vendor claims unless your lawyer says otherwise.

What if the vendor offers to pay for our data?

Then it is a buyer, and the usual comparisons apply. Google, which also sells Google Workspace, runs one such program. 404 Media reported on June 2, 2026 that Google had emailed Play Store developers offering to pay for their app code. The pilot’s own page calls it non-exclusive and says partners keep their IP; the profile covers how the price is set, which starts with a number you propose.

An offer from a company you rely on every day still needs the same answers in writing: which records, for how long, exclusive or not, and paid when. The offer benchmark lays one out term by term.

What should I check?

For each system that holds your records:

1. The current terms and AI or privacy page, and the date you read them.
2. The clause on who owns your content, and any clause on generated, usage or aggregate data.
3. The license you gave the vendor and its permitted uses, including AI or machine learning.
4. Whether any training serves your account only, features every customer gets, or other companies’ models.
5. The opt-out or setting, its default, and who holds the admin role that controls it.
6. Any limit on using exports or API access for AI training, which Is it legal? and the QuickBooks Online and Slack pages set out.

What do I do next?

1. Decide on each opt-out. It is a separate decision from licensing. Slack’s global models, Atlassian’s in-app data on Free and Standard and ServiceTitan’s data preferences stay on unless someone turns them off.
2. Add the vendor terms to your data inventory, one line per system with the date checked, since a buyer may ask what each vendor can do with the records.
3. Give counsel those terms with any draft license, especially an exclusive one.
4. If the vendor has made an offer, lay it out in the offer benchmark and apply for others to compare it with.