A quant fund's edge lives in a model nobody outside the desk has ever seen. That is the entire point of the model. The moment you register it with a patent office, describe it in a copyright filing, or hand a full specification to a court exhibit, the edge is gone, whether or not you win the case. This is the practical problem that sits underneath every dispute about who actually built a trading strategy first: the standard tools for proving ownership all require you to show your work, and showing your work is the one thing a trading strategy cannot survive.
Why patents and copyright do not fit a trading strategy
A patent protects a novel, non-obvious invention, but only after the application is published. Publication is the mechanism, not a side effect. A strategy filed as a patent is a strategy any competitor with a database subscription can read in full. Most quantitative strategies also fail the patentability bar in the first place, since abstract mathematical methods and business logic are routinely excluded or narrowed to the point of uselessness during examination.
Copyright is closer but still the wrong tool. Copyright protects the specific expression of an idea, the literal text of your code or the wording of a research memo, not the underlying method, formula, or trading logic itself. Two developers can independently write completely different code that implements the same signal, and copyright will not stop either of them. If your dispute is about who thought of the approach first, copyright in a codebase answers a different question than the one being asked.
What actually gets disputed
The disputes that show up in practice are rarely about copying code line for line. They fall into a small number of recurring shapes:
- A quant leaves the desk and starts a new fund running a strategy that looks close enough to the old one that origination becomes the whole argument.
- A joint development or advisory arrangement ends, and both sides claim they built the core model, with the other side treated as a later contributor.
- A counterparty who saw a strategy under NDA during due diligence later runs something similar, and the fund has to show what the counterparty actually had access to and when.
- An internal researcher's early draft gets picked up by another team, and nobody can agree on which version came first or how complete it was.
In every one of these, the question a court or arbitrator actually needs answered is narrow: who had a complete, specific version of this model, and on what date. Not what the model says. Just that it existed, in that exact form, at that exact time, in someone's possession.
Proof of possession without disclosure
This is a sealing problem, not a registration problem. The mechanism does not require the strategy to be readable by anyone reviewing the proof, including the fund's own lawyers preparing a filing. A cryptographic hash is a fixed-length fingerprint of a file. Change one character anywhere in the strategy document, model weights, or code repository and the hash changes completely. The hash itself reveals nothing about the content it was generated from, which is exactly the property a fund needs: you can prove you possessed a specific, unaltered file at a specific moment without ever putting that file, or even a description of it, in front of anyone.
Pair that hash with a qualified electronic timestamp from an accredited trust service and you get a dated, tamper-evident record. Under the EU's eIDAS Regulation, Article 41 gives a qualified electronic timestamp the presumption of accuracy for the date and time it indicates and the integrity of the data it covers, which shifts the burden of proof onto whoever wants to challenge it rather than onto the fund holding the seal. Switzerland's own certification framework, ZertES, extends comparable legal weight to qualified signatures and time evidence issued by accredited Swiss providers. Neither of these frameworks asks what the underlying document says. Both only certify that a specific, unmodified artifact existed at a specific point in time.
What this looks like in a real dispute
When a departing quant claims independent development, the fund's position is only as strong as its dated evidence. A well-timed seal of the complete strategy specification, model code, and backtest results, created before the departure and never altered afterward, gives counsel something concrete to point to: a specific version, cryptographically fingerprinted, timestamped by an independent trust service, months or years before the competing claim surfaced. That is a materially stronger starting position than an email thread or a Git commit history that a skilled opposing expert can characterize as ambiguous or reconstructable.
The same logic applies to counterparty disputes during fundraising or partnership due diligence. Sealing the version of the strategy shared under NDA, at the moment it was shared, creates an independent record of exactly what was disclosed and when, separate from whatever either side later claims was discussed in a meeting.
Building this into how a desk already works
The practical discipline is simple and does not require changing how research is done: seal a snapshot of the model, its documentation, and its backtest artifacts at defined milestones, not only at the end. A useful cadence is at first working version, at every material revision that changes trading logic, and immediately before any disclosure to an outside party, whether that is a counterparty, an advisor, or a departing employee's final handover. Each seal is a small, unglamorous administrative step that produces a dated fingerprint the fund may never need. When a dispute does arrive, that fingerprint is often the difference between a credible claim of origination and a hard-to-prove assertion.
Before assuming a strategy carries limited exposure, it is worth running the numbers on what a dispute over origination would actually cost in legal fees, lost time, and reputational damage with investors. The IP exposure calculator walks through that estimate for a specific strategy or model in a few minutes.
