Adobe in Hot Water: Faces Class-Action Lawsuit Over Alleged Copyright Misuse in AI Training

On December 17, 2025, a significant legal battle emerged in the tech world as a proposed class‑action lawsuit was filed against Adobe in the U.S. District Court for the Northern District of California. The company is accused of infringing copyright laws by allegedly misusing copyrighted books, including those authored by Elizabeth Lyon, in the training of its AI model, SlimLM. This case underscores the increasing tension between AI development and intellectual property rights, marking Adobe's entry into ongoing legal challenges faced by tech giants like Apple, Microsoft, and Salesforce over similar allegations involving AI training data sourced from copyrighted materials. The lawsuit highlights the contentious use of the SlimPajama‑627B dataset, claimed to derive from the RedPajama dataset's Books3 collection, sparking debate over the limits of AI innovation and content creators' rights.

As Adobe steps into the spotlight with its AI model, SlimLM, the class‑action lawsuit draws attention to the broader challenges facing tech companies in ethically sourcing training data for AI applications. According to TechCrunch, this legal move not only impacts Adobe but also serves as a bellwether for future AI developments and the protection of intellectual property. SlimLM, developed for mobile document assistance, utilizes the open‑source SlimPajama‑627B dataset, yet the case negotiated by McGuire Law PC contends that the dataset comprises unauthorized copyrighted content. This forms part of a broader narrative where datasets such as Books3 are at the center of legal scrutiny, challenging the norms of data usage in AI's persistent growth.

The lawsuit against Adobe, filed in the U.S. District Court for the Northern District of California, marks a significant legal battle in the realm of artificial intelligence and copyright laws. Alleged in the suit is that Adobe has misused copyrighted materials to train its SlimLM AI model. SlimLM's development allegedly involved the SlimPajama‑627B dataset, which contains unauthorized versions of copyrighted books, including works by the plaintiff, Elizabeth Lyon, a nonfiction author from Oregon. According to the,¹ this lawsuit represents Adobe's first major confrontation with copyright infringement issues in AI, a challenge that other tech giants like Apple, Microsoft, and Salesforce are already facing.

Adobe's AI model known as SlimLM is designed specifically to enhance document assistance tasks on mobile devices. While the company boasts of the model's capabilities, it also faces scrutiny due to the manner in which SlimLM was reportedly trained. According to a,¹ it is alleged that the training involved unauthorized use of copyrighted content sourced from the SlimPajama‑627B dataset. This dataset, in turn, is said to incorporate elements from the controversial Books3 collection, part of the RedPajama dataset, which comprises a considerable number of copyrighted books.

SlimLM represents Adobe's strategic efforts in advancing AI technology tailored for mobile utilities, enabling users to perform a wide range of document‑related tasks efficiently. The model's development hinged on training data from the open‑source SlimPajama‑627B dataset, which has drawn legal challenges due to claims of containing copyrighted material. As echoed in the lawsuit initiated by nonfiction author Elizabeth Lyon, there's rising concern about the ethical implications of using such data without proper authorization, posing questions about Adobe's data governance practices.

Essentially, the SlimLM model is part of Adobe's broader initiative to pioneer small language models that operate seamlessly within mobile environments. Despite the technical achievements, the lawsuit raises crucial discussions on the legality and ethics of utilizing datasets like SlimPajama‑627B, which is accused of integrating content from contentious sources such as the Books3 collection. This situation places SlimLM at the center of an intense debate regarding copyright issues, fair use, and the permissible scope of AI training datasets, as reflected by industry observers.

The connection between Books3 and RedPajama plays a pivotal role in the ongoing legal challenges faced by Adobe. Books3, a collection of approximately 191,000 books, serves as a significant component of the RedPajama dataset, which has been under scrutiny for including potentially unauthorized copies of copyrighted material. Adobe's AI model, SlimLM, has come under fire for allegedly utilizing the SlimPajama‑627B dataset, which is claimed to be a derivative of RedPajama's Books3 collection. This linkage highlights the intricate web of data sources that fuel AI training and the crucial intersection between intellectual property rights and technology advancements, as demonstrated in the lawsuit discussed.¹

Within the lawsuit against Adobe, the SlimPajama dataset is a central point of contention. As a derivative of RedPajama, it illustrates the controversial blending of datasets that include copyrighted works from Books3. The amalgamation of such datasets raises questions about the legality of using them to train AI models like SlimLM.¹ This case underscores the broader industry challenge of balancing innovation with respect for creator copyrights, a tension reflected in similar lawsuits faced by tech giants such as Apple and Microsoft, which have also utilized the Books3 corpus for AI training purposes.

Previous similar lawsuits involving AI and intellectual property rights have highlighted the contentious nature of using copyrighted material in training datasets. Companies like Apple have faced legal challenges over their AI models being trained on the same Books3 database that is central to many claims, including the one against Adobe. In September 2024, Apple was sued by authors who alleged that Apple Intelligence improperly utilized the Books3 dataset, which consisted of thousands of copyrighted works used without authorization in their machine learning models. This legal scrutiny suggests a broader industry pattern, as Adobe's situation mirrors the increasing tension between technological advancement and copyright law as companies push the boundaries of AI capabilities while grappling with existing legal frameworks.

The case against Anthropic in 2024 stands out as another pivotal moment in the legal landscape of AI training. Anthropic, facing allegations similar to those now targeting Adobe, opted to settle with a consortium of authors for a significant sum, reportedly around $1.5 billion. This settlement was perceived as a landmark moment, laying the groundwork for authors to leverage similar claims against other tech companies. It underscored the potential financial risk tech companies take on when using datasets like Books3 without clear authorization from rights holders.

Salesforce has also been implicated in lawsuits concerning the use of the Books3 and RedPajama datasets. Allegations surfaced that Salesforce's AI models utilized these datasets, which included derivative works from copyrighted materials. This case exemplifies the legal challenges faced by corporations using complex data collections that often lack transparency concerning the intellectual property rights of included content. Much like Adobe, Salesforce's legal troubles illustrate the pressing need for clear legal guidance on AI development and intellectual property management.

The trend of lawsuits, including multiple federal class actions, spotlights the increasingly critical issue of dataset provenance in AI training. Companies like Microsoft and other major AI vendors have been brought into the legal spotlight, with courts tasked to consider whether using such comprehensive book datasets constitutes an infringement or falls under fair use. These ongoing litigations are crucial for setting legal precedents, which might influence how future datasets are compiled, potentially demanding more stringent adherence to copyright laws.

In response to the surge of legal actions, there has been a growing focus within the tech industry on improving the documentation and transparency of datasets used in AI training. This process often includes efforts to deduplicate content and implement takedown procedures for identifiable copyrighted material. For instance, the SlimPajama dataset, used by Adobe, is part of a broader trend in which dataset maintainers attempt to adapt their practices in line with legal expectations and avoid future litigations. This proactive approach reflects a shift among technology companies aiming to mitigate risks by ensuring their training data complies with legal and ethical standards.

The public reactions to the Adobe lawsuit regarding its AI model, SlimLM, have sparked a widespread and polarized debate across various platforms. On social media channels like X (formerly Twitter), there's a significant wave of support for the plaintiff, Elizabeth Lyon, and other authors who claim their works were used without permission. Contributors to Reddit forums have also echoed these sentiments, criticizing Adobe for allegedly exploiting creators to enhance AI capabilities without proper compensation. This viewpoint is encapsulated in various viral tweets and posts highlighting the perceived hypocrisy of Adobe—an enterprise traditionally known for creative software—being embroiled in a case of alleged copyright infringement, as detailed in a recent.¹

Conversely, a substantial segment of the tech community defends Adobe, positing that the use of the SlimPajama‑627B dataset and the AI model's aims are within legal boundaries. This group argues under the fair use doctrine, noting that open‑source datasets should allow innovation in AI technology. Comparisons have been drawn to past legal precedents such as Authors Guild v. Google, which supported the notion of fair use in similar scenarios. Discussions on Hacker News and other forums have detailed technical aspects like dataset vetting and transformation to advocate for Adobe's approach, while some financial analysts on platforms like TradingView consider the lawsuit insignificant against the backdrop of Adobe's robust market performance, as observed in their latest earnings report.

The debate extends beyond traditional tech circles, touching on broader ethical and societal concerns. On LinkedIn and other professional networking sites, mixed reactions reveal a divide where some view the legal action as an essential step in protecting authors' rights in the digital age, while others fear that such lawsuits might hinder technological advancement and the democratization of AI tools. As noted in eWeek coverage, the implications of this lawsuit may lead to significant shifts in how AI training datasets are composed, potentially enforcing more rigorous documentation and transparency when deploying AI technologies.

Amidst these discussions, a recurring theme is the comparison with Apple's prior legal challenges related to the Books3 dataset. The Adobe case is often placed within a larger narrative of 'AI copyright wars,' where tech giants are increasingly scrutinized for their data usage practices. This situation invites frequent comparisons and speculations about potential outcomes similar to the substantial settlements observed in similar lawsuits against companies like Anthropic, as reported by Law360. As the lawsuit progresses, it is poised to set new legal precedents and perhaps reshape the landscape of AI training, copyright law, and creative intellectual property rights.

The recent lawsuit against Adobe, accusing it of copyright infringement due to its use of unauthorized works in training its SlimLM AI model, stands to have significant implications for the AI industry. If the case results in a legal precedent, AI companies might need to reassess their data curation processes, potentially leading to stricter licensing and increased costs for ethically sourcing training data. This could stifle innovation, especially among smaller AI developers who may not afford the high price tags associated with compliant datasets. Such a trend might hike the cost of developing AI models, particularly compact ones intended for specific tasks like SlimLM's document assistance capabilities, thereby dampening the industry’s rapid growth trajectory.

The Adobe lawsuit, among others targeting companies like Apple and Microsoft for similar infringements, underscores a growing tension between AI advancements and copyright law. As the industry grapples with these legal challenges, companies might begin prioritizing transparency in their data training methodologies to mitigate litigation risks. The outcome of these cases could compel AI developers to move towards 'author‑approved' datasets and intensify the scrutiny of open‑source datasets like SlimPajama‑627B. Furthermore, it could drive legislative efforts to establish clearer guidelines for what constitutes fair use in AI training — a move that might redefine the boundaries of innovation and intellectual property in the digital age.

In the wake of the class‑action lawsuit filed against Adobe, there are several possible pathways the legal outcomes could take. If the court sides with Elizabeth Lyon, the plaintiff, it could result in substantial financial settlements for unauthorized use of copyrighted material in AI model training. This might include damages, as well as injunctions to prevent further usage of the dataset in question. Such rulings could set a significant precedent, influencing other similar cases involving major tech companies like Apple, Microsoft, and Salesforce. As noted in previous instances, such as the Anthropic settlement, financial restitutions could potentially reach billions, impacting Adobe and similar organizations financially and operationally.

However, if Adobe is able to successfully argue that the use of Books3‑derived materials falls under the fair use doctrine, it might shield itself from liabilities and any compensatory payments. The company might emphasize the transformative nature of its AI application or challenge the characterization of specific dataset components as infringing material. The outcome will likely hinge on a detailed analysis of how these principles apply in the context of AI model training. This approach is not unprecedented, as seen in cases like Authors Guild v. Google, where fair use was successfully argued concerning digital reproduction for a different technological application.

Apart from the immediate financial implications, the lawsuit could drive legislative and regulatory reforms around AI and intellectual property rights. Current debates within the industry highlight a growing consensus on the need for clearer guidelines on how copyrighted content can be used in data training models. If Adobe and other tech giants are pushed towards settlement, this might catalyze broader shifts towards more ethical data governance practices, including mechanisms for compensating content creators whose work is utilized in AI training, akin to the settlement models observed in other significant AI copyright cases.

Additionally, regardless of the court's ruling, this lawsuit could propel changes within the industry itself. Companies might begin investing more heavily in synthetic data or ensuring explicit licensing agreements, moving away from potentially infringing datasets. This was a trend highlighted in market analyses following RedPajama‑related lawsuits, where firms increasingly scrutinized their data sourcing methods to avoid legal entanglements and public relations pitfalls. As seen in Adobe's history of innovation, this could potentially lead to the development of new tools or methodologies that align more closely with evolving legal landscapes and public expectations.

The legal battle between Adobe and Elizabeth Lyon underscores a pivotal moment in the AI industry, highlighting the tension between technological innovation and intellectual property rights. As the lawsuit unfolds, it presents a significant test case for how copyright laws are interpreted and applied in the context of AI training data. This confrontation could potentially reshape the landscape, compelling companies like Adobe to reassess their data sourcing and training practices to ensure compliance with existing legal frameworks.

This case is not just about Adobe; it's emblematic of a broader struggle within the tech industry, where the rapid advancement of AI models often outpaces the legal systems in place to govern them. Similar lawsuits against other major tech giants like Apple and Microsoft indicate a growing trend where authors and creators are taking a stand to protect their intellectual property rights in the digital age. If the courts decide in favor of Lyon and other authors, it could set a legal precedent that requires tech companies to obtain explicit permission before utilizing copyrighted works for AI training.

Moreover, the outcome of this lawsuit could influence public perception and legislative actions regarding AI development. If Adobe is found culpable, it may catalyze stricter regulations and force companies to adopt more transparent and ethical data collection methods. This might involve developing more robust opt‑out mechanisms for authors and negotiating fair compensation for using their works, potentially leading to broader industry reforms enhancing the rights of content creators in the face of advancing AI technologies.

The societal implications are profound, with potential shifts in how AI models are developed and accessed. Authors like Lyon champion the cause for greater respect and valuation of human creativity, arguing that unauthorized use of their work for AI training devalues their original content and unfairly benefits tech companies. This issue resonates with the broader public, fostering a dialogue about the ethical use of technology and its impact on culture and creativity. As tech companies navigate these challenges, the decisions made today could shape the landscape of AI and copyright for years to come.