🧠 Expertise Snapshot
An AI business can combine a foundation model, proprietary data, retrieval tools, and a carefully designed user experience. Each layer raises a different IP question. At Miller IP Law, our approach starts by separating the technology you create from the platforms, models, and content you license.
For AI and large language model businesses, that means examining the actual system: how information enters, how relevant context is retrieved, how outputs are checked, and how software acts on the result. This technical focus helps connect patent, trademark, and copyright decisions to the parts of the business that create value. A chatbot demo is the introduction. The architecture is where the deeper conversation starts.
⚙️ How Our Expertise Makes a Difference
A useful AI patent discussion goes beyond “we added an LLM.” We examine what happens between the user's request and the final response: document processing, retrieval, model selection, tool permissions, validation, and fallback behavior. The goal is to identify a specific technical contribution and explain how it works.
Consider a system that retrieves records according to user permissions, checks a generated answer against those records, and blocks unsupported actions. Our analysis would examine those interactions, the available prior art, and the evidence of improvement. The same business also needs a brand strategy and clear ownership of its code. Coordinating these decisions helps founders prioritize protection around what customers actually buy.
🔒 IP Breakdown: Protecting What Matters
Patents may protect qualifying technical inventions. Trademarks protect names and other source identifiers. Copyright protects original human expression, including qualifying code and documentation. Trade secrets and contracts can complement these rights for confidential methods, data access, and ownership.
🔬 Patents in AI and LLM Businesses
Putting an existing business task through an LLM does not automatically create a patentable invention. A stronger evaluation examines the claimed technical solution, such as a particular retrieval process, memory arrangement, or inference control mechanism, and how it improves computer operation or another technical field.
We focus on describing the mechanism, alternatives, and supported technical benefits rather than promising a result. Patent eligibility is only one requirement; novelty, nonobviousness, and adequate disclosure also matter. AI assistance does not eliminate the need to identify the human inventor or inventors. U.S. inventorship turns on human conception under the ordinary legal standard, even when AI tools assist development.
🏷️ Trademarks in AI and LLM Businesses
Your product name may become more durable than the model powering it. Trademark planning starts with a distinctive brand and a search for potentially conflicting marks. A domain purchase or available app store name does not establish that a mark is clear. Descriptive names can also be difficult to register and protect.
The application must identify the actual goods or services. Downloadable AI software commonly falls in Class 9, while hosted software commonly falls in Class 42; the correct description depends on the offering. An AI education platform may raise different coverage questions. We connect the filing scope to what the business provides, so “AI platform” does not become a substitute for describing the product.
📚 Copyrights in AI and LLM Businesses
Copyright can protect original human-authored code, documentation, and other expressive material in an AI product. It does not grant ownership of the underlying idea or function. For AI-assisted work, the central question is which expressive elements a human actually authored. Keeping records of creative revisions can help establish that contribution.
Under U.S. Copyright Office guidance, prompts alone generally do not establish authorship of generated output. Qualifying human edits or creative selection and arrangement may be protected. We distinguish those contributions from generated material when evaluating registration scope. Permission to use an output under a vendor contract is a separate question from whether that output has copyright protection.
💡 Real-World Business Examples
Microsoft: Brand Identity
Microsoft's published trademark guidelines identify product names, logos, and icons as proprietary brand assets and set rules for their use. For an AI startup, the practical lesson is that recognizable branding can remain valuable while the software behind a product changes.
Adobe: Contracts and Creative Tools
Adobe offers contractual IP indemnification for select Firefly outputs under qualifying arrangements, subject to conditions and exclusions. That protection addresses specified third-party claims; it does not establish copyright ownership in every generated output. The example shows why AI businesses should evaluate vendor terms alongside their own IP rights.
These public examples illustrate different protection tools, not client engagements. Miller IP Law is not identified as handling either company's IP. For a growing AI business, the takeaway is to coordinate brand protection, ownership analysis, and licensing before scaling distribution.
🎯 Take the Next Step
Building an AI or LLM business? Let's identify what you created, what you license, and which IP protections fit your goals. Schedule a consultation at strategymeeting.com
