# IP Strategy for Deep Tech Startups A patent portfolio is either a moat or vapor. There is no middle ground. The filing hierarchy: - **Filed PCT with national phase entries:** Real. Costs real money. Claims can be evaluated. - **Filed but not yet published:** Promising, but claim scope unknown until publication. - **Developed but not filed:** Dangerous. If it's your most defensible IP, why haven't you filed? Every month of delay is a month a competitor could file first. - **Conceptualized:** Not IP. It's a slide. The two most important claimed capabilities being "conceptualized" rather than filed is a red flag that should block any round above seed. The industrial AI IP paradox: the components (knowledge graphs, surrogate models, RL optimization) are all standard and unpatentable. The integration and orchestration layer (automated pipeline from raw data to deployed model) is potentially novel and defensible, but companies often fail to patent it because it's "just engineering." Broad titles are a warning sign. "Methods and Systems for Asset Management" could claim anything or nothing. Until you read the actual claims in the WO publication, you don't know what's protected. The competitive clock: in industrial AI, incumbents can replicate any unpatented capability in 6-12 months. A startup with no filed patents on its core claimed capabilities has a 12-18 month window before the moat disappears entirely. Related: [[Technical Moat Assessment Framework]], [[Technical DD Framework]], [[Industrial AI MOC]] --- Tags: #deeptech #investing