Adobe clarified its updated terms of use, but the update initially led users to believe the unpublished work was material for AI training.
Adobe says so do not have The company states that it trains its Firefly generative AI models on user content and that it “never assumes ownership of its customers' work.”
Backlash surged this week after Adobe users were notified of updates to its terms of service. One line that caught attention was in section 2.2: “Our automated systems may use techniques, such as machine learning, to analyze Your Content and Creative Cloud Customer Fonts (defined below in Section 3.10 (Creative Cloud Customer Fonts)) to improve our Services and Software and your user experience.”
The tweet may have been deleted
Adobe clarified in a blog post that this part of the policy is not new, and is specifically focused on moderating illegal content, such as child sexual abuse content and content that violates their terms, such as creating spam or phishing attempts. “Given the explosion of generative AI and our commitment to responsible innovation, we have added additional human moderation to our content submission review process,” the blog post read. In other words, they've added additional human moderation, rather than automated moderation.
Mashable Lightspeed
But Adobe didn't indicate exactly what was updated in its terms of service, leading users to interpret the clause to mean that unpublished work, including sensitive content, could be used to train AI models. “Our promise to our customers remains unchanged,” the blog post read. Adobe explained that Firefly models are trained on licensed content, such as Adobe Stock, and public domain content. Adobe also argued that any work users create in the app is their property.
The tweet may have been deleted
What exactly did Adobe update to its Terms of Use?
In the blog post, the company highlighted in pink text where the wording has been changed. In the Privacy section, the phrase “only access” has been changed to “may access, view, or view your content” and the line “through both automated and manual methods, but only in limited ways” has been added. Later in this paragraph, Adobe adds the line “as further set forth in Section 4.1 below.” Section 4.1 outlines the details of content that is illegal or prohibited. The paragraph also adds the phrase “including manual review,” which refers to the methods employed to screen Adobe's content. In this section, “child pornography” has been changed to “child sexual abuse material.”
This means that Adobe’s privacy policy remains unchanged, which may come as a relief to users who rely on Adobe for their creative endeavors. But the reaction to Adobe’s vaguely worded update reflects creators’ fear and distrust that generative AI could upend their professions. AI models such as ChatGPT, Dall-E, Gemini, Copilot, and Midjourney have been criticized for training on content gleaned from the web to automate writing and image creation, and OpenAI’s Sora, which is not publicly available, is believed to be trained on videos from YouTube and elsewhere.
Overall, it's a tense time for creative people, and they have reason to be wary of changes that could threaten their livelihoods.
topic
Artificial Intelligence Privacy
