# IP Encumbrance Map ## The company says that it holds patents, trade secrets and know-how, but no one can examine them The company describes IP protection as a combination of patents, trade secrets, contractual confidentiality and internal know-how. The company operates for some years. The company says that it applied for patents. The company does not identify one patent by number, title, jurisdiction or assignee. The company protects its methods for the work and the tuning of the model as trade secrets. This method is reasonable and it ties value to team retention. ## Encumbrances to map before a reader relies on the IP | Source of encumbrance | Why it can matter | | --- | --- | | Academic source of the first work | The IP policies of a university or a national research council can give those bodies rights in first methods or models. | | Government-funded projects | The terms of the project can have IP conditions or publication conditions. | | Client-owned outputs | The largest contract assigns the models that the company trains and the validation data to the client. The files have no background-IP schedule. | | Derived-output restrictions | One customer or more do not let the company use the outputs and insights of the customer again. | | Customer exclusivity | The company cannot put exclusive agents for a customer back into the platform layer. | | Third-party components | The license terms of open-source models and foundation models apply. See [[Open-Source Dependencies and Licence Exposure]]. | > [!warning] The company gives no proof of the patent claims > "Applications filed" without numbers is a claim and not an asset. Green means that the assignee of specified methods is the company. For example, the methods are knowledge capture or root-cause analysis in a factory. Red means one of two things. The first is that the company applied for no patent. The second is that the assignee is a person or a university. ## Why it matters IP in a small applied-AI company is mostly contract terms and personnel. The applicable diligence has two parts. The first part is a matrix of data rights and background IP for each vertical. The second part is a founder IP assignment. The company gives the two parts to legal diligence. Compare this with the framework in [[IP Strategy for Deep Tech Startups]]. ## Related - [[Foreman MOC]] - [[Borrowed Data Is Not a Moat]] - [[A short note on IP]]