Contentstack now keeps your data out of public AI training

The change

On June 5, 2026, Contentstack replaced its old Supplementary Terms with a rewritten Artificial Intelligence Addendum. Venpo detected the rewrite and flagged it for review.

This is not the kind of change we usually write about. Most vendor updates take something away from the customer. This one mostly gives something back. Contentstack used the rewrite to strengthen how customer data is protected when its AI features are used, and to remove several clauses that previously favored the vendor.

Contentstack is a headless CMS that sits at the content layer of many digital products. As it rolls out AI tools like Agent OS, Agent Builder, AI Assistant, and Brand Kit, the terms governing what happens to the data flowing through those tools matter to every customer relying on them.

What changed in customers' favor

The strongest addition is a clear restriction on AI training:

• Contentstack will not use customer data to develop, train, or improve any general or public AI features

• Contentstack does not permit its AI providers to use customer data to train or improve their own public AI or foundational models

That second point is the one most teams miss. It is not enough for a vendor to promise it won't train on your data; the AI providers sitting behind that vendor have to be held to the same line. Contentstack now does both in writing.

Three more changes move in the customer's direction:

• Inputs and outputs, including proprietary brand data fed into Knowledge Vaults, are explicitly defined as Customer Data owned by the customer

• Several provider-favorable clauses were removed, including a right to terminate AI features with no notice and no refund, and the customer's obligation to indemnify Contentstack for a broad range of AI-related claims

• A standalone carve-out that had stripped all warranty and SLA protection from AI services is no longer stated in the same blanket way

What also changed, in fairness

A balanced reading means naming the parts that add obligations, not just the wins. Three changes lean the other way, though none are aggressive:

• The software is now expressly not for use as an EU AI Act "high-risk AI system," and customers are prohibited from deploying it that way. Most customers are not building high-risk systems, so this affects a narrow set of use cases.

• Customers must now maintain human oversight, review, and validation of AI output, including for AI agents and autonomous workflows. This is closer to a statement of good practice than a new burden.

• The definition broadened from a narrow "AI Integration" to "AI Features," pulling more named tools under these terms.

On balance, the rewrite is clearly pro-customer. But the only way to weigh the wins against the obligations is to see the full change, not a headline.

Why this matters

AI data handling is one of the most scrutinized parts of any vendor relationship right now, and the trend has mostly run against customers. We have written about vendors moving in the opposite direction, from Anthropic introducing a clause that can override zero-data-retention commitments to analytics tools changing their AI data defaults. Against that backdrop, a vendor tightening its own AI data protections is genuinely notable.

It also matters because positive changes are easy to miss entirely. A negative change might eventually surface through a complaint, an incident, or a customer questionnaire. A favorable change usually surfaces through nothing at all. If you are not watching the document, you simply never learn that your vendor got safer, and you lose the chance to act on it. This is the same visibility gap covered in our analysis of the hidden risk of vendor legal changes, just pointed in the other direction.

Potential impact for SaaS companies

Companies using Contentstack's AI features may want to:

• update their vendor risk records to reflect the new no-training commitment, which may strengthen their own compliance position

• use the clarified Customer Data ownership language when answering their own customers' security questionnaires

• note the removed indemnity and termination clauses, which reduce previously one-sided risk

• confirm none of their use cases fall under the new high-risk AI prohibition

A favorable vendor change can be an asset, not just a relief. The no-training commitment is the kind of fact that helps you reassure your own customers, and it only works in your favor if you know it exists. Tracking changes across AI providers in your stack is exactly where monitoring AI subprocessors and broader continuous vendor risk monitoring pay off in both directions.

How Venpo detected it

Venpo continuously monitors vendor legal documents, regardless of whether a change helps or hurts the customer. When Contentstack replaced its terms, Venpo:

• detected that the Supplementary Terms had been rewritten into a new AI Addendum

• identified the new no-training commitment covering both Contentstack and its AI providers

• flagged the favorable ownership, indemnity, and termination changes alongside the new obligations

• separated the genuine wins from the added requirements so the net effect was clear

An objective monitor reports what actually changed, not only the alarming changes. That is the difference between monitoring for fear and monitoring for awareness, and it is the argument at the center of our comparison of manual vs automated vendor monitoring.

Business outcome

Companies that caught this change early were able to:

• record a stronger AI data protection posture for one of their vendors

• improve their own customer-facing answers about how data is handled downstream

• retire previously flagged concerns about AI indemnity and no-notice termination

• do all of this proactively, rather than discovering the improvement by accident months later

Instead of operating on outdated assumptions about a vendor's AI terms, they updated their picture the moment it improved. Current information is an advantage whether the news is good or bad.

Key takeaway

Not every vendor change is bad news. Contentstack rewrote its AI terms to keep customer data out of public AI training, confirm that inputs and outputs belong to the customer, and remove several clauses that favored the vendor. It is a real improvement, and a reminder that monitoring is not only about catching threats. The only way to know your vendor got safer is to be watching the document when it changes, in both directions. The case for catching every change, not just the damaging ones, is laid out in our analysis of manual vs automated vendor monitoring.