Adobe recently faced backlash from users after an update to its Terms of Use policy sparked concerns about the use of unpublished work for AI training. However, the company has now clarified its position and stated that it does not train its Firefly generative AI models on user content, nor does it assume ownership of customer work.
The specific language in Section 2.2 of Adobe’s Terms of Use policy caused the controversy. The sentence stated that Adobe’s automated systems may analyze user content and Creative Cloud Customer Fonts using techniques like machine learning to improve their services and software. This led users to believe that their unpublished and confidential work could potentially be used for AI training.
To address these concerns, Adobe released a blog post explaining that the language in question was not new and was primarily focused on moderating illegal content and violations of their terms. Adobe clarified that it has added more human moderation to its content submissions review processes, rather than relying solely on automated moderation.
The company emphasized that its Firefly models are trained on licensed content, including Adobe Stock, and public domain content. Additionally, Adobe reiterated that users retain ownership over the work they create using its apps, emphasizing that its commitment to customers has not changed.
To provide further clarity, Adobe highlighted the specific changes made in its Terms of Service. In the privacy section, the language was modified from “‘will only’ access” to “‘may’ access, view, or listen to your Content.” The addition of the phrase “through both automated and manual methods, but only in limited ways…” was also made, with reference to Section 4.1 for further details.
The updated Terms also included the addition of “including manual review” to describe the methods employed for screening Adobe’s content. Furthermore, Adobe replaced the term “child pornography” with “child sexual abuse material” in line with more appropriate terminology.
While these updates do not substantially change Adobe’s privacy policy, the reaction to the initially vague language highlights the fear and mistrust felt by creatives in the face of generative AI. AI models like ChatGPT, Dall-E, Gemini, Copilot, and Midjourney have faced criticism for being trained on web-scraped content, enabling automated writing and image creation. The unreleased OpenAI model, Sora, is believed to be trained on videos from platforms like YouTube.
This tense environment has created a sense of caution among creatives who fear that changes in AI technology may threaten their livelihoods. The concerns raised by Adobe’s update reflect the broader anxieties within the creative community as they navigate the impact of AI on their professions.
In conclusion, Adobe has clarified its Terms of Use update and assured users that their unpublished work is not used for AI training. The company’s commitment to customer ownership and its emphasis on responsible innovation provide reassurance to creatives who rely on Adobe’s tools for their creative pursuits. However, the overall reaction to the update underscores the unease and caution that many creatives feel as AI continues to advance.
