By Víctor M. Rodríguez-Reyes, Senior Member Attorney at Ferraiuoli
In 2026, the intersection of technology and intellectual property law continues to evolve at breakneck speed. The start of 2026 delivered critical developments that will define the legal landscape for years to come.
“Here’s Johnny!”, would be a fitting quote to describe AI’s forceful entrance on the legal scene, though, as I will discuss, it has not been without rebuff. AI’s ubiquitous and pervasive comingling of data and its rapid adoption is having immediate, transformative effects on the daily practice of law, often creating a fine line between a helpful tool and a significant liability.
The Battle for Control: Who owns what in an AI world?
Federal Courts Reject AI as Counsel: Recent rulings have made it clear that AI cannot serve as legal counsel. Federal judges have dismissed lawsuits filed by “artificial entities” attempting to use AI for representation, characterizing the practice as the unauthorized practice of law. These courts observed that AI-driven arguments often result in “incomplete arguments directed toward unrelated issues,” highlighting the current gap between AI potential and courtroom readiness.
Pro se AI not excused: The risks are even higher for self-represented litigants. In Wilcox v. Gingrinch, a pro se brief was found to contain fourteen completely fabricated cases generated by AI. The court emphasized that pro se litigants are “held to the same standards as a trained attorney” and are afforded no leniency for submitting fictitious AI-generated authorities. Consequently, their primary legal arguments were waived because the foundational authority did not exist.
NYC: Client-Side AI Tools Create Privilege Risks: Even outside the courtroom, AI use creates potentially unsuspected issues. NYC Ethics Opinion 2025-6 warns that when clients use their own AI tools to record or transcribe conversations with their lawyers, the risk they may waive privilege may increase, because this shifts control from a lawyer bound by professional duties to a client who may be unaware of the risks of unreviewed AI summaries.
The “Whistle Blower” Risk: USA v. Heppner, perhaps the most startling development is the potential for AI to act as a de facto whistle blower against its own users. In this criminal case (February 17, 2026), the court ruled that a defendant’s communications with the AI platform “Claude” were not protected by attorney-client privilege or the work product doctrine.
The “morale of the story” is that information introduced into an AI system can be discoverable by the opposing party. If AI must be used, use enterprise-quality systems that explicitly state in their policies that user data is not used for training, though users must remain vigilant as these policies change regularly.
In some contexts, the use of AI will result in obvious copyright infringement. A potential example, includes Hollywood studios—including Disney, Paramount, and Warner Bros.—alleging against ByteDance, the creator of the AI video generator Seedance 2.0, “pervasive copyright infringement,” arguing that ByteDance used their copyrighted films as “raw material” for training without consent, where the original works are clearly identifiable in the output.
However, in other AI contexts, proving copyright claims is much harder due to techniques like distillation. Anthropic recently accused laboratories such as DeepSeek and Moonshot of using 24,000 fake accounts to siphon capabilities from its Claude model.
Distillation involves a “student” model learning from a “teacher” model’s outputs rather than raw data. This creates several legal hurdles:
- No Copyright Ownership: You cannot claim rights that do not belong to you. For example, if a company’s terms assign output ownership to the user, the AI lab itself may not hold the copyrights necessary to sue for extraction.
- Lack of Human Authorship: The U.S. Copyright Office has affirmed that copyright requires human authorship; therefore, products produced entirely by AI are not protectable.
- The Public Data Dilemma: If data is public, it is unlikely somebody can claim ownership over it. Therefore, there is debate over whether “publicly available” means accessible to everyone for any purpose, or if internet service providers can legally lock down data they have indexed for search engine purposes.
- The Cohere Problem (RAG-Enabled Copying): A similar issue exists with Cohere’s Retrieval Augmented Generation (RAG) feature. When RAG accesses external data sources in real-time, outputs allegedly include verbatim copies, substantial excerpts, and “substitutive summaries” mirroring expressive choices and narrative structure, not just facts. A landmark ruling in Advance Local Media v. Cohere found that these summaries can infringe copyright if they parrot the organization and style of the original work rather than just reporting facts.
Remedies and the Preemption Trap: Currently, companies seek remedies through legal actions citing violations of terms of use and website policies. However, these state-law claims often face the hurdle of copyright preemption. In ML Genius Holdings, LLC v. Google LLC, the Second Circuit dismissed breach of contract claims because they were “coextensive” with rights protected by the Copyright Act, effectively barring state-law claims for copying.
To avoid this preemption problem, platforms are adopting new strategies. Reddit is challenging AI scrapers by invoking the DMCA’s anti-circumvention provisions (Section 1201), alleging they bypassed Google’s technological measures to access Reddit’s content. Similarly, Google has sued SerpApi, claiming its scraping methods circumvent security measures. By focusing on “access controls” rather than just the content itself, these companies hope to use the DMCA to bypass traditional copyright preemption.
Conclusion: The Evolving Landscape
The latest months have delivered transformative developments across technology and intellectual property law. Key themes emerge:
- IP Rights Apply to AI: Courts make clear copyright and trademark law fully apply to AI systems, including novel theories addressing hallucinations and substitutive summaries.
- Specificity Matters: Whether trade secrets, service contracts, or copyright claims, detailed documentation and contractual provisions are essential.
- Ethics Evolve with Technology: Legal professionals must adapt ethical frameworks to AI-enhanced practice, understanding both opportunities and risks.
- For businesses: in the short run, institutions and employers need to manage these risks by instituting policies that address AI uses, implement best practices, and protect their assets. In the long run, monitoring regulatory developments closely, investing in IP protection early and comprehensively, document everything, seeking specialized counsel, and staying informed.
- For legal professionals: Update technical knowledge, review evolving ethics guidance, advise clients proactively, and think strategically about navigating overlapping federal and state requirements.
The intersection of technology and law has always been dynamic, but the pace of change in 2025-2026 is unprecedented. Staying informed, engaging expert counsel, and thinking strategically about intellectual property protection aren’t optional, they’re essential for success in the modern innovation economy.
FAQ’s
What is a “substitutive summary” in copyright law?
A substitutive summary is an AI-generated output that, while not a verbatim copy, mirrors the expressive choices, narrative structure, and journalistic storytelling of the original work. The Cohere decision holds these can infringe copyright when they go “well beyond a limited recitation of facts” by “lifting expression directly or parroting the piece’s organization, writing style, and punctuation.”
