Adobe Reiterates It Does Not Train Firefly AI on Customer Content

After a mini-flareup over a revised Terms of Use agreement, Adobe has clarified that it will not use customer content to train its AI, nor will it in any circumstance claim ownership of customer content that was created with its AI tools.

“We recently made an update to our Terms of Use with the goal of providing more clarity on a few specific areas and pushed a routine re-acceptance of those terms to Adobe Creative Cloud and Document Cloud customers,” an Adobe statement reads. “We have received a number of questions resulting from this update and want to provide some clarity. We remain committed to transparency, protecting the rights of creators, and enabling our customers to do their best work.”

Adobe is semi-unique in the generative AI space because of its rich history, enormous user base of professional creators, and its responsible AI pledges. And so what seemed like a rote update to its Terms of Use ignited a firestorm of controversy, similar to the silliness over Microsoft Recall, except that the Adobe policies concern shipping products and services that impact millions of users.

More specifically, Adobe’s previous Terms of Use update noted that the firm “may access your content through both automated and manual methods, such as for content review, but only in limited ways, and only as permitted by law.” As The Register pointed out, this isn’t new, it’s been Adobe’s policy for several years, and it predates the current AI push.

Regardless, Adobe is answering the feedback with further adjustments to its Terms of Use that clarify the conditions under which it would need to “collect, use, share, or otherwise process” customer content. And it’s added a new clause to address this in more detail.

“We reserve the right (but do not have the obligation) to remove Content or restrict access to Content, Services, and Software if any of your Content is found to be in violation of the Terms,” a new Section 4.1 reads. “We do not review all Content uploaded to the Services and Software, but we may use available technologies, vendors, or processes, including manual review, to screen for certain types of illegal content (for example, child sexual abuse material) or other abusive content or behavior (for example, patterns of activity that indicate spam or phishing, or keywords that indicate adult content has been posted outside of the adult wall).”

More explicitly, Adobe has reiterated two core customer promises:

• Adobe does not train Firefly Gen AI models on customer content. Firefly generative AI models are trained on a dataset of licensed content, such as Adobe Stock, and public domain content where copyright has expired.
• Adobe will never assume ownership of a customer’s work. Adobe hosts content to enable customers to use our applications and services. Customers own their content, and Adobe does not assume any ownership of customer work.

You can learn more at Adobe Help.