Counsel that crosses the whole product
Ownership, licensing, and risk are not separate filings — they run through the model, the data, and the terms together. We advise the product as one system, not a stack of disconnected documents.
PRACTICE · AI LAW
AI & Machine Learning
Cross-cutting counsel for the people actually building and shipping AI and machine-learning products — not a single document, but the IP, licensing, and risk questions that run through a venture from first model to first customer. We help founders and operators decide what they can own, what they have licensed, and where the exposure sits, in a body of law that is unsettled and still moving.
- Discipline
- AI Law
- Engagement
- Product counsel
- Counsel
- Christopher Moye
AI LAW
Counsel that crosses the whole product
Ownership, licensing, and risk are not separate filings — they run through the model, the data, and the terms together.
The problem
Most AI ventures ship a product before anyone has settled what they own, what they licensed, and who answers when the model is wrong.
The questions cut across the whole company — authorship of outputs, the provenance of training data, the vendor terms behind the model, the ownership of fine-tuned weights and prompts. Each is governed by law that is unsettled and moving, and each surfaces at a different moment: a funding round, a customer's diligence, a takedown notice. The counsel that helps is practical and cross-cutting — built into the product now, not assembled after a claim arrives.
Principles · 01
How we draft the matter.
Every engagement is composed against these commitments. They shape the protections we add, the questions we ask, and the document that leaves the file.
Build the record before the claim
Authorship, data provenance, and ownership are decided by what you can show. We help structure the development and contracting records now, so the position is documented before a customer's diligence or a takedown asks for it.
Advise to the law as it stands
The questions around AI are unsettled and moving. We do not claim a certainty the courts have not provided; we build defensible positions that hold up across the ways the open questions may resolve.
What we watch · 02
What can break the matter.
These are the terms, structures, and practical risks that usually decide whether the work holds when the file is tested.
FOUNDERCTO
Ownership of outputs and models
What your product can claim in its generated outputs and in the fine-tuned weights, pipelines, and prompts your team builds — documented so the position survives diligence.
FOUNDERDATA OFFICER
Training data and licensing
Where your data comes from, what rights travel with it, and how the licenses and source agreements behind the product allocate infringement risk.
OPERATORGC
Vendor terms and liability
The model and vendor terms your product depends on, and the allocation of risk — in those terms and in your own — for when a model errs.
The work · 03
Four steps. One engagement.
Each step is concrete; each step has a deliverable. The scope is defined, the matter moves, and the file closes.
- 01
Product review
We map the product as it is built — the models it calls, the data it holds, and the terms it ships under — to locate where ownership and risk actually live.
- 02
Rights and exposure mapping
We trace what you own, what you have licensed, and where infringement and liability exposure sits across the model, the data, and the contracts.
- 03
Documents and terms
We draft and negotiate the ownership, licensing, and risk terms — and the records behind them — so the positions are in writing rather than assumed.
- 04
Ongoing counsel
We revisit the work as the product, the contracts, and the law change, and bring in the firm's deeper AI specialties where a matter calls for them.
Proof
What stands behind the work.
What stands behind the work — credentials and representative engagements, stated plainly.
AI and machine-learning matters are handled by Christopher Moyé, Esq., who authors the firm's published writing on AI law, authorship, and the ownership of models and their outputs.
Product-stage counsel for AI and ML ventures — IP ownership of outputs and fine-tuned models, training-data and licensing review, vendor and model terms, and allocation of risk when a model errs. The firm's deeper specialties — IP strategy, governance, procurement, and policy — feed into this work where a matter calls for them.
We start from the product as it is built — the models it calls, the data it holds, the terms it ships under — and advise to the current state of the law rather than a settled answer the courts have not given.
Common questions
Questions clients ask.
Plain answers to the questions that come up most. If yours is not here, send the facts — we answer in writing.
- Can our company own the copyright in what our AI product generates?
- Only to the extent of human authorship. The U.S. Copyright Office has refused registration for purely machine-generated output, and the courts that have reached the question — including the Thaler line — have held that copyright requires a human author. Where your team contributes copyrightable expression, a claim can rest on that contribution, and we structure and document the human role so it is visible. The standard is still developing, so we advise to the current rule and do not promise a registration will issue.
- Could the data our model was trained on expose us to infringement?
- It can. Pending cases concern models trained on copyrighted material and outputs that resemble protected works, and the law on training data and fair use is unsettled. We review where your training data came from and what rights you hold in it, read the licenses that govern any data you bring in, and allocate infringement risk in the agreements with the vendors and sources behind the product. We do not guarantee compliance or a particular outcome — the questions are live and the answers are moving.
- If our model produces something false or harmful, who is liable?
- Liability turns on the contracts and the facts, not on a single rule. Model vendors routinely disclaim responsibility for output; your customer agreements and disclaimers shift or retain it from there. We allocate that risk across the chain — in the vendor terms, in your own terms of service, and in the use restrictions and review checkpoints around the product — so an error is governed by a deliberate allocation rather than a default. We cannot promise a model will not err or that a court will read the allocation as written.
- Who owns the fine-tuned weights and the prompts we developed?
- It depends on what the underlying model's license permits and what your contracts say. A base model usually remains the provider's, but the fine-tuned weights, the training pipeline, and the prompt systems your team builds can be your proprietary assets — protected as trade secrets and, in part, by contract — if the agreements are drafted to keep them yours. We read the model license against your development records and structure the terms so ownership of what you built is not surrendered to the provider by default.
- What should we negotiate in the terms for the models we build on?
- The terms that decide whether your inputs stay yours, whether the vendor trains on your data, who owns the output, and who bears an infringement claim. Default model agreements often let the provider retain and train on submitted data and disclaim liability for the output. These are negotiable, and the procurement review is where that protection is won — a deeper specialty the firm handles in its own right, which we bring in when a model's terms sit at the center of your product.
FOR THE FOUNDERS WHO SHIP
Counsel for what you are building.
Bring the product as it actually exists — the models you call, the data you hold, the terms you ship under. We map what you own, what you have licensed, and where the risk sits, and put the documents in order.
Begin a conversation