Adobe is updating its terms of service to clarify how customers' data is used to train its AI models. Adobe will introduce the revised terms next week.
The company aims to rebuild trust by explicitly stating that it will not utilize customers' work for AI training purposes.
Adobe is updating its terms of service to clarify how customers' data is used to train its AI models.Justin Sullivan/Getty Images
Adobe Clarifying the Terms of Service
Adobe's upcoming release of terms of service, scheduled on June 18, aims to provide clearer guidelines on how the company can utilize its customers' work.
This decision follows concerns raised by users over a recent update to Adobe's terms of service, prompting the company to address these issues through a blog post announcement.
David Wadhwani, the president of digital media at Adobe, emphasized their commitment to never training generative AI on customer content, never claiming ownership of customer work, and never granting access to customer content beyond legal obligations.
Over the past week, Adobe encountered significant criticism from its user base following the discovery of ambiguous language regarding AI in its updated terms of service. Many customers interpreted this vague wording as indicating that Adobe intended to utilize their work freely to train its generative AI models.
However, this assumption was inaccurate, as Adobe's policies on training remained unchanged. Adobe's chief product officer, Scott Belsky, admitted that the wording was unclear and emphasized the importance of trust and transparency in today's landscape.
Wadhwani clarified that the wording in Adobe's Terms of Service was never intended to authorize AI training on customers' work. He admitted that, in hindsight, they should have updated and clarified the terms earlier, taking proactive steps to align them with their actual practices and provide more precise explanations of their legal obligations.
Many in the creative community have ongoing concerns about Adobe, citing its perceived industry dominance, subscription-based pricing, and use of generative AI.
While Adobe aimed to address ethical issues by training its Firefly AI model on openly licensed and public domain content from Adobe Stock images, some artists have discovered images resembling their work on the platform, raising doubts about the effectiveness of existing safeguards.
Wadhwani expressed confidence in Adobe's content moderation process for Adobe Stock and Firefly training data but admitted it's not flawless. He emphasized that Adobe can eliminate content breaching its policies from Firefly's training data. He noted that customers have the option to opt out of automated systems aimed at enhancing the company's services.
In its blog post, Adobe emphasized the importance of earning trust, saying it values feedback to address upcoming changes. While increased transparency is a positive step, winning back the trust of disappointed creatives may require some time.
Adobe expressed its commitment to becoming a trusted partner for creators in the future and pledged to exert continuous effort toward that goal.
