Allen Adamson is Co-Founder of Metaforce, Adjunct Professor at NYU Stern School of Business, and an expert witness in trademark, branding, and consumer perception disputes.
This column examines how Meta’s AI training practices could erode trademark distinctiveness, increase brand confusion, and create new legal risks for professional service firms and trademark attorneys.
Originally published July 14, 2025 in Law.com.
A recent reporting in The Atlantic has highlighted a new and urgent dimension to the risks posed by large language models. While much of the public’s attention has focused on copyright infringement, an equally important concern has received far less discussion: the threat that AI training poses to trademark distinctiveness, particularly for professional service brands.
While developing their Llama 3 model, the article reveals that Meta employees debated whether to license copyrighted content legally or simply pirate it. Facing high licensing costs and delays, they downloaded millions of books from the pirated Library Genesis database. Internal communications show that Meta employees acknowledged this presented a “medium high legal risk” and discussed steps to conceal the source of the pirated material.
While the copyright issues are obvious, the trademark risks are equally serious. When AI models train on unlicensed professional content, thought leadership, proprietary frameworks, and strategic methodologies, they absorb the distinctive expressions that brands rely on to differentiate themselves in the marketplace. These expressions are often critical elements that trademark law protects through the concept of distinctiveness.
Trademark law is designed to protect a brand’s ability to serve as a source identifier. The more distinctive a brand’s presentation, voice, and strategic content, the stronger its trademark protection. When unauthorized AI training weakens or copies these distinctive brand signals, firms face real threats to their brand identity.
For professional service firms in fields such as consulting, investment management, legal services, and marketing, brand value is built not only on credentials but on the proprietary thinking they bring to clients. This intellectual property, expressed through white papers, published articles, conference presentations, and signature frameworks, creates the brand’s voice and establishes its competitive position. It is these unique elements that trademark law helps protect from misuse or confusion in the marketplace.
If AI models ingest distinctive strategic content and begin generating similar advice or frameworks, the risk of dilution grows. Distinctiveness can erode. Trademark law recognizes several harms that could arise from this dynamic.
First, there is the loss of distinctiveness, where a brand’s signature voice becomes so commonplace that it no longer serves as an effective source identifier. Second, there is the risk of confusion if AI outputs resemble a firm’s voice or proprietary materials so closely that clients cannot distinguish the origin. Third, there is trademark dilution, particularly through blurring, where the uniqueness and strength of a brand are gradually weakened by imitation, even in the absence of direct competition.
The Atlantic article makes clear that this is not a hypothetical risk. Meta’s internal communications show an awareness that training on pirated materials could create problems beyond copyright. One employee warned that torrenting pirated materials from a corporate laptop “does not feel right.” Others discussed ways to strip metadata, including ISBNs and copyright notices, to obscure the original sources.
These actions suggest a broader recognition that using pirated professional content could have lasting impacts on brand owners, including those whose intellectual property defines their distinctiveness in the market.
While current litigation focuses primarily on copyright claims, trademark implications are likely to grow more prominent as AI models advance. Creating lookalike content that blurs the lines between one firm’s intellectual expression and another’s can undermine the basic function that trademarks are intended to serve.
If clients or prospective clients cannot recognize the source of a firm’s strategic insights because similar content has been absorbed into AI models and distributed widely, the brand’s hard-won differentiation may be permanently eroded.
Professional service firms, and the trademark attorneys who advise them, should take this risk seriously. Several steps are worth considering now:
Conduct a thorough audit of proprietary intellectual property, especially content that expresses strategic frameworks or brand voice.
Evaluate publication and distribution strategies to limit uncontrolled exposure of proprietary materials where appropriate.
Monitor the marketplace for AI-generated outputs that resemble the firm’s distinctive brand elements.
Consider aligning with broader industry coalitions taking action against unauthorized AI training practices.
Educate internal stakeholders on how brand distinctiveness could be impacted by unauthorized AI use and the importance of reinforcing trademarks with clear, consistent brand signals.
Protecting trademark distinctiveness has always been important for service brands, but it becomes even more critical in an environment where generative AI can blur the lines between original thought and machine-generated imitation. Unauthorized use of proprietary materials to train AI models not only appropriates content but also jeopardizes the very competitive advantage that strong trademarks represent.
As AI products become more powerful, the gap between a firm’s authentic brand identity and a machine generated approximation will continue to narrow unless firms take action. Clients choose professional service brands not just for solutions, but for the distinctive quality of insight, judgment, and strategic creativity they represent. Protecting that distinctiveness is no longer just a matter of good brand management. It is a matter of trademark survival.
Trademark law has always safeguarded a brand’s ability to signal a unique origin and promise of quality. In the age of AI, maintaining that distinctiveness will require new vigilance, new strategies, and a renewed commitment to protecting what makes each brand different, valuable, and trusted.
