An agency built our MVP. Do we own it?

Only if they signed an assignment. Work made for hire is narrower than most founders assume and generally covers employees, not independent contractors or agencies. Without a written assignment the developer may hold copyright in code you paid for in full, leaving you with an implied licence to use it and nothing more. It is straightforward to fix while the relationship is good, and considerably harder once it is not.

Do we own what our AI feature generates?

Not the machine generated portion. United States copyright requires human authorship, and the Copyright Office has been consistent that purely machine generated output is not protectable. What is protectable is the human contribution: selection, arrangement, editing and creative direction. In practice this means structuring the workflow so there is a documented human element, and being careful about what your terms promise customers regarding ownership of output.

Is our software patentable?

Sometimes, and the honest answer is often no. Software patents face eligibility challenges that mechanical inventions do not, and a patent with claims a competitor can trivially design around has cost you money for very little. For many software businesses the defensible moat is trade secret discipline, execution speed and data, not a filing. Where a genuine technical improvement exists, it is worth pursuing properly. Where it does not, we will say so.

Industries

Software, platforms and AI

SaaS, marketplaces, developer tools, AI products and anything with an API. The recurring problem is not what you built, it is proving you own it when someone finally asks.

Book a Screening Call All industries

What makes this sector different

Software companies rarely fail diligence on their technology. They fail on paperwork. A contractor who never signed an assignment, a copyleft licence three levels deep in the dependency tree, terms of service copied from a competitor that describe a business you do not run: each of these is cheap to fix in advance and expensive to discover during an acquisition.

AI has added a second layer. What a model was trained on, whether the output is ownable, and what your own terms say about customer data are now standard questions in enterprise procurement. Answering them before a customer asks is faster and considerably cheaper than answering them in a security questionnaire.

Typical matters

Contractor and developer IP assignment

Getting written assignments in place, and repairing the chain of title when code was written before anyone thought to ask for one.

Terms of service and privacy policy

Enforceable user obligations, warranties, service levels and liability limits written for how your product is actually sold, not for a generic business.

Open source licence audit

Finding copyleft and attribution obligations in your dependency tree before an enterprise customer or an acquirer runs the same scan.

AI ownership and training data risk

What you can own in machine assisted output, what your prompts and training data expose you to, and what your terms should say about both.

Software and platform patents

Where a software invention is genuinely patentable, claims drafted to survive the eligibility challenges this field attracts, and an honest answer when it is not worth filing.

API, data sharing and integration terms

Terms for developer access, plug ins, datasets and analytics, including who may derive what from whose data.

Services used most in this sector

A printed agreement marked up in red pen with sticky tabs

Questions from this sector

An agency built our MVP. Do we own it?
Only if they signed an assignment. Work made for hire is narrower than most founders assume and generally covers employees, not independent contractors or agencies. Without a written assignment the developer may hold copyright in code you paid for in full, leaving you with an implied licence to use it and nothing more. It is straightforward to fix while the relationship is good, and considerably harder once it is not.
Do we own what our AI feature generates?
Not the machine generated portion. United States copyright requires human authorship, and the Copyright Office has been consistent that purely machine generated output is not protectable. What is protectable is the human contribution: selection, arrangement, editing and creative direction. In practice this means structuring the workflow so there is a documented human element, and being careful about what your terms promise customers regarding ownership of output.
Is our software patentable?
Sometimes, and the honest answer is often no. Software patents face eligibility challenges that mechanical inventions do not, and a patent with claims a competitor can trivially design around has cost you money for very little. For many software businesses the defensible moat is trade secret discipline, execution speed and data, not a filing. Where a genuine technical improvement exists, it is worth pursuing properly. Where it does not, we will say so.

Start with a 30 minute consultation.

A $50 video call covering your goals, your timeline and the documents you send ahead. The $50 is credited toward your fee when the firm takes your matter on.

Book a Screening Call Call 512-612-6454

Software, platforms and AI

What makes this sector different