The change

Contentful is a content platform that companies use to manage and publish the content behind their websites and apps, and Salesforce completed its acquisition of Contentful on September 1, 2026. On November 9, 2026, Contentful published a new version of its AI Terms of Service, the terms you accept before you turn on its AI features. The version it replaces was effective May 3, 2026.

On the same day, its subprocessor list added a new entry for Datadog, a monitoring service, that covers the same AI content.

What changed

• A broad promise became a narrow one. Contentful no longer promises not to use your AI content to develop or improve its service. It now promises only not to train generative AI models on it without your opt-in.

• New permitted uses. Contentful may use your AI content to improve the services and features you have access to, and for research and development.

• An open-ended third use. It may also use that content “As otherwise provided in your agreement with Contentful.”

• A new subprocessor. Datadog may now process AI content to monitor, test and improve Contentful’s AI features.

What “Customer AI Content” means

The terms define Customer AI Content as “(a) any input you provide to be processed through the Contentful AI Functionality and (b) any output generated and returned by the Contentful AI Functionality based on your input.”

In practice that is the text, content and instructions your team sends to Contentful’s AI features, and everything those features send back.

The old promise

Section 4.2.2 covers paid customers, those whose use “is governed by a Master Subscription Services Agreement or other written subscription agreement with Contentful”. The May version said Contentful “will not use your Customer AI Content to develop or improve the Subscription Services, including to train or otherwise improve any underlying foundation models”.

It allowed two exceptions: “(i) where this is done solely to provide the Contentful AI Functionality to you and for your sole benefit, or (ii) with your authorized representative’s written consent.” So any improvement had to be for your own benefit, or you had to agree in writing.

The new promise

The November version says Contentful “will not use your Customer AI Content to train any underlying generative AI models unless you provide affirmative opt-in consent.”

It then adds: “Contentful may access Customer AI Content for the following purposes and you instruct Contentful to process Customer AI Content for such purposes:”

• “To improve the services and features to which you have access;”

• “To conduct research and development of services and features to which you will have access without additional cost; and”

• “As otherwise provided in your agreement with Contentful.”

The difference is in what is left out. Training generative AI models still needs your opt-in. Improving the product and doing research no longer need anything, and they are no longer limited to your sole benefit. The words “you instruct Contentful” frame these uses as your own instruction, which is how data processing contracts describe what a vendor is allowed to do.

Two smaller edits point the same way. The section title lost the words “for Training”, and the ban now names “generative AI models” where it used to say “foundation models”. The new opt-in wording also does not say the consent must be written or come from an authorized representative.

Datadog joins the list

The subprocessor list now includes Datadog, Inc., based in the United States, with Germany listed as the data location. Its listed purpose is “Processing Customer AI Content in connection with certain features that use Contentful AI Functionality, including agentic features, to monitor service quality, validate product changes, troubleshoot issues, and improve those features.”

That matches the new terms. Improving AI features is now an allowed use of your AI content, and Datadog is one of the companies that may handle it for that purpose.

What did not change

Contentful’s outside AI providers are still barred from training on your content. Section 4.2.4 says they will have “contractual restrictions in place with Contentful prohibiting use of Customer AI Content to train or otherwise improve any underlying generative AI models.”

Free and self-serve accounts were never covered by the paid promise. Section 4.2.3 says Contentful “may use your Customer AI Content to develop and improve the Subscription Services, including through training and fine-tuning of our AI models.” Contentful’s AI FAQ also still says: “No, Contentful does not use your organization’s data to train the models that support AI Actions.” That answer is about training, so it does not cover product improvement.

When the new terms apply to you

Section 2.1 says: “Contentful will provide you four (4) weeks’ notice of any material revisions by posting notice within the Contentful Subscription Services web portal, on its website, and/or via email to the administrative email address associated with your account.” It also says: “Your continued access or use of the Contentful AI Functionality constitutes your acceptance of any revisions.”

Your way out is in the same section: “If you object to any revised Terms, you can disable the Contentful AI Functionality.” If your own contract links to a specific version of the AI terms, that version applies instead.

Why this matters

AI promises are being rewritten across the SaaS stack. Retool dropped its promise not to train its own AI on customer data, Mux’s new terms let it train its own AI models on customer data, and Articulate’s AI training promise now covers only generative models. Asana can now use your data to improve its product without aggregating it, too.

Contentful’s version keeps the training promise and widens everything around it. A promise not to train sounds complete, but product improvement and research are where your content is now allowed to go. New AI providers on the subprocessor list are part of the same picture, which is why it helps to monitor AI subprocessors in your vendor stack.

Potential impact

If you use Contentful, check five things:

• Does your team use Contentful’s AI features, and on a paid plan with a written agreement?

• What kind of content goes into those features, and would you mind if it were used to improve Contentful’s product?

• Do your own customer contracts or privacy notices limit how your vendors may use that content?

• Does your Contentful contract link to a specific version of the AI terms?

• Did anyone at your company receive or read the notice of this change?

How Venpo detected it

Venpo monitors Contentful’s legal pages and compared the November 9 versions line by line with the earlier ones on record. The details are on the Contentful AI Terms change page and the Contentful subprocessor change page. Every quote in this article was checked against Contentful’s live pages and the May 3, 2026 version of the AI terms at the time of publication.

Business outcome

Teams that catch this can decide on purpose whether their AI content goes into Contentful’s product work, update their own notices, or turn the AI features off. Teams that miss it accept the new uses by simply using the features. That is the case for continuous vendor risk monitoring and for continuous monitoring of vendor legal changes.

Key takeaway

Since November 9, 2026, Contentful’s AI terms let it use paid customers’ AI prompts and outputs to improve its services and for research, keep only a promise not to train generative AI models without opt-in, and add Datadog as a processor of that content.