Data Ownership in Cloud Contracts: CAUCUS Conversation Summary

Presented by Brian Walsh, Chair of CAUCUS

The cloud makes everything easier, right? Upload your data, access it anywhere, scale on demand. What could go wrong?

Everything. Everything can go wrong—and usually does—because most organizations have no idea what's actually happening with their data once it hits the cloud. In a recent CAUCUS Conversation, Brian Walsh pulled back the curtain on one of procurement's most dangerous blind spots: data ownership in cloud contracts.

The Data You Don't Know You're Handing Over

Most buyers think "data" means the customer database they uploaded or the files sitting in their cloud storage. But that's where the manipulation starts.

Your data actually includes:

  • Customer names, addresses, transactions

  • Metadata and system configurations

  • Derivative works (forms, automations, anything you build with their tools)

  • Proprietary business processes and trade secrets

  • PII, HIPAA data, classified information, financial records

Think about that last one. If you're using Adobe AEM to build customer intake forms, the form itself is your data. The automation behind it? Your data. The submissions? Your data. And if your contract doesn't explicitly say otherwise, your vendor might be using all of it to train their next AI model.

The Geographic Shell Game

Here's a favorite vendor trick: You negotiate to keep your data in the United States. You get it in writing. You feel good. You're protected, right?

Wrong. If your data can be accessed from outside the US, it's not actually US-only data. When their overseas support team logs in to troubleshoot your problem, your supposedly US-based data just became global data.

What vendors do:

  • Promise data will be "housed in the United States"

  • Never mention their global support teams can access it anywhere

  • Charge you incrementally more for each geographic restriction layer

  • Use vague terms like "region" (North America ≠ United States)

Brian's advice: Your contract needs to specify both where data is stored and where it can be accessed from. Include audit language that lets you check transaction logs. And yes, you'll pay more for these restrictions—but at least you'll know what you're actually getting.

The DPA Trap

Every major cloud provider has data protection buried in something called a DPA—Data Processing Agreement or Data Protection Addendum. And here's the problem: it's always a link.

Brian was emphatic: "Links in contracts are evil." Why? Because they change. That link in your contract might say one thing today, but in two years when the vendor updates their DPA, that same link suddenly says something completely different. And buried in their service terms is language that says they can change those linked agreements at any time.

What to do instead:

  • Take the language from the linked DPA and make it an exhibit to your contract

  • Include the date and time the DPA was active

  • Lock down the specific version you're agreeing to

Otherwise, you have no idea what you actually agreed to—and the vendor can change the rules anytime they want.

Real EULAs, Real Problems

Brian walked through actual end-user license agreements from major vendors (names changed to protect the guilty). The findings were damning.

SillyMonkey: "It's Your Data (But We're Using It)"

SillyMonkey's EULA clearly states: All data submitted through surveys belongs to you. Great! Except keep reading: They can use your data "to improve existing services and develop new services." 

Translation: Your customer survey responses—the data you paid to collect—is training their AI. And when your contract ends? They can hold your data for two years. If it includes trade secrets, they can keep it forever (as long as the trade secret remains a trade secret).

ZWS: The Region Shuffle

ZWS lets you choose which "region" your data lives in—North America, Europe, Asia. Sounds good until you read the fine print in their supply chain DPA:

"You agree and instruct that we may use and store your content that is processed by the supply chain service to develop and improve the service and its underlying technologies."

And this gem: "We may store your content in a region outside of your region where you're using the service." In other words, you picked North America, but they're moving your data wherever they need it for their AI training. You can opt out—if you find the opt-out policy buried in the administrative guide.

Minisoft: The Only One That Gets It Right

In a rare moment of honesty, Minioft's EULA repeatedly states: You own your data. They have no rights to it. They won't touch it. It's refreshingly clear—and probably why Brian called out their language as the model to follow.

Getting Your Data Back (Before It's Too Late)

You've been using Salesforce for years. You finally realize you're paying for a CRM built for enterprises when you're a 50-person company. Time to switch. How do you get your data out?

If your contract says "Contractor must return all relevant user data upon termination," congratulations—you're about to get a completely unusable CSV file. Brian has seen this happen. Vendors meet the letter of the contract while delivering something worthless.

What your contract needs:

  • Specific format for data return (not "mutually agreed upon")

  • Exact timeline (60 days, not "upon termination")

  • Vendor assistance requirement if extraction is complex

  • Attestation that all data has been deleted from their systems

  • Verification process to confirm data integrity

And start the extraction process before your contract ends. Once you're no longer paying them, you're dead to them. Good luck getting help after termination.

The AI Training Ground You're Paying For

Can vendors use your data? Brian's answer: "The short answer is no. Your default answer should always be no. The long answer is absolutely not."

But here's reality: Unless your contract explicitly prohibits data scraping, AI training, machine learning, third-party sharing, and aggregation for analytics, they're probably doing all of it. It's how they build better products. It's how they train AI models. It's how they create competitive advantages—using your data.

Your contract must explicitly prohibit:

  • Data mining and analysis for vendor purposes

  • Aggregation across customers

  • Marketing and product development using your data

  • AI and machine learning training

  • Third-party sharing (including affiliate companies)

Don't rely on vague statements like "we keep your data confidential." Define what they can't do, specifically and exhaustively.

Breach Notifications: The Timeline Trap

Your organization requires breach notification within 24 hours so you can notify customers, credit bureaus, regulators. Makes sense. But your cloud vendor's contract says they'll notify you within seven days of discovering a breach.

See the problem? All your internal compliance obligations just became impossible to meet. Your contracts need to align. If you must notify in 24 hours, your vendor must notify you in time to meet that deadline.

Brian shared an example from his workers' comp days: Their standard contract was 10 pages. Their data contract? 40 pages. The data clause alone was 30 pages—with 5-6 pages just defining what data meant. Why? Because the risk was monumental, and they needed ironclad protections.

Sample Clauses (As Starting Points, Not Final Language)

Brian used Microsoft Copilot to compile data protection language from state contracts. The results were generic but useful as starting points. Here's what needs to be in your contracts:

Data Ownership

"All data, metadata, and derivative works provided or generated under this agreement remain the exclusive property of [Company]. Contractor shall have no rights, title, or interest in such data except as expressly required to perform services under this agreement."

Add: Explicit prohibition on data scraping, AI training, machine learning, and third-party sharing.

Geographic Restrictions

"All data must be stored within [specify country/state]. All access to data must originate from within [specify country/state]. Contractor must provide transaction logs for audit upon request."

Data Exit

"Upon termination, Contractor must return all Company data in [specify exact format] within [60] days. Contractor must provide written attestation that all data has been permanently deleted from all systems, including backups. Contractor must cooperate to ensure complete and accurate data transfer."

Encryption

"Data must be encrypted at rest and in transit using [specify standard, e.g., AES-256]. Contractor must notify Company within 24 hours of discovering any breach, including root cause analysis and remediation plan."

Prohibition on Vendor Use

"Contractor may not mine, analyze, aggregate, or repurpose Company data for any purpose other than fulfilling obligations under this agreement. This includes prohibition on use for marketing, product development, AI training, machine learning, or analytics."

The CAUCUS Standard

This session exposed what CAUCUS members already know: Vendors are built to extract maximum value from you while providing minimum transparency. Data contracts are where this gets dangerous because the stakes are existential—a breach, a misused dataset, a disappeared backup can destroy organizations.

The vendor playbook is predictable:

  • Hide critical terms in linked DPAs that can change anytime

  • Use vague language ("region" not "country," "upon termination" not "60 days")

  • Bury data usage rights in page 300 of supplemental documents

  • Make geographic and access restrictions incrementally expensive

  • Train AI on your data unless you explicitly opt out

Your defense is specificity. Define everything. Lock down every variable. Embed linked documents as exhibits. Specify formats, timelines, encryption standards, access restrictions, deletion requirements. Leave nothing to interpretation.

The Questions You Need to Ask

Brian emphasized these aren't statements—they're questions you must answer for every contract involving data:

  • What is our data? (All of it—base data, metadata, derivative works)

  • Where does it live? (Specific country/state, not just "region")

  • Who has access? (Where can they access it from?)

  • How is it protected? (Encryption at rest and in transit)

  • What's our exit strategy? (Exact format, timeline, deletion attestation)

  • Can the vendor use our data? (Default answer: No. Explicit answer: Hell no.)

Final Word

"If it's not written in the contract, it's not real," Brian reminded us. "Links can change. Verbal promises disappear. Only what's explicitly in your signed agreement matters."

Most organizations treat cloud contracts like hosting agreements when they're actually data custody agreements. The difference matters. You wouldn't hand your most valuable assets to a third party without ironclad controls on what they can do with them. Your data deserves the same protection.

Upcoming CAUCUS Events:

  • February 19: AI Indemnification - Who Pays When the AI Screws Up?

  • March 4: Termination for Convenience - Building Your Exit Before You Need It

Keep the conversation in the #Vendor Free Forum Your data is too important to leave to vendor goodwill.

Full recording and slides are now available in #Conversations