Every AI product is built on someone's model, and each model comes with terms that shape what you can ship. We review open-weight licences, hosted model provider terms and the DPAs that sit with them, and license your own models out on terms you choose. Many open-weight licences are not standard open-source licences, so we read the actual conditions rather than assume.
At a glance
- Practice
- AI Infrastructure
- Pricing
- Flat fee, quoted before we start
- Included
- 6 deliverables
- Process
- 3 steps, one named attorney
- Where we work
- Slack, email or a call
When founders call us
The moments that usually bring a company to us for this work.
- 01
Your product is built on an open-weight model and you want to know whether its licence permits your commercial use.
- 02
You fine-tuned a base model and need to understand what obligations carry over to your derivative.
- 03
An enterprise customer asked whether your model provider can train on their data.
- 04
You are releasing your own model weights and need to choose or draft a licence.
What you get
Scoped in writing before we start. One flat fee for all of it.
- Review of open-weight licences for commercial limits, user thresholds, acceptable-use policies and attribution or naming requirements
- Analysis of obligations on fine-tuned and derivative models and on outputs
- Review of hosted model provider terms, including data use, retention and zero-retention options
- DPA review or negotiation with model providers
- Outbound licence drafting for your own models or weights
- Flow-down checklist so your customer terms reflect upstream model obligations
How it works
- 1
Inventory your models
We list every model you use or ship and the licence or terms attached.
- 2
Map the obligations
We identify the restrictions, attribution and flow-down duties that apply to your product.
- 3
Fix and align
We update your customer terms or negotiate provider terms to close the gaps.
Mistakes we help you avoid
The ones we see most often, and catch before they cost you.
- Avoided
Assuming 'open weights' means unrestricted commercial use when the licence has use limits or a user threshold.
- Avoided
Shipping a fine-tuned model without the attribution, naming or acceptable-use flow-down the base licence requires.
- Avoided
Promising customers their data is never used for training while your model provider's default terms say otherwise.
- Avoided
Licensing your own weights without restrictions you later wish you had.
Model Licensing: questions founders ask
Not necessarily. Some are released under permissive open-source licences, while others use custom licences with acceptable-use policies, attribution requirements or limits for very large users. You need to read the specific licence for the specific version you use.
Many custom model licences extend their conditions to derivatives, which often includes fine-tuned versions. That can mean passing through acceptable-use terms, keeping notices or following naming rules. Whether and how a licence reaches your model depends on its wording.
It depends on the provider's terms, your plan tier and any settings or agreements you have in place. Many providers offer enterprise or API terms that exclude training on inputs, but defaults vary. We review the terms you are actually on and align your customer promises with them.
If you send personal data to the provider, you generally need a DPA to meet GDPR and many state privacy law requirements. Most major providers offer one, and we review whether it covers your use case and matches what you tell your customers.
Get a flat-fee quote for Model Licensing.
A short call with an attorney. You leave with a written scope and one fixed price.
Book a call