Prior Art Under Pressure - The Diligent Search in UPC Litigation

The Unified Patent Court (UPC) has created a fast and consequential forum for European patent disputes. A defendant generally has only three months after service of an infringement claim to submit its defense and any counterclaim for revocation. Within that period, the legal team must develop its invalidity positions, locate prior art, assess the evidence and prepare the pleading.

This front-loaded process creates an understandable demand for speed. As Benedikt Biechele of Global Prior Art explains in this article, AI-enabled tools can help teams respond by identifying patents containing similar language, tracing citations, suggesting terminology and ranking potentially relevant documents. Keyword searching remains equally useful for testing concepts and exploring known vocabulary. The more important question is how these capabilities should fit within a search designed for high-stakes litigation.

A diligent search aims to maximize both precision and recall. Precision measures the percentage of identified references that are relevant to the search focus. Recall measures the percentage of all relevant references that a search successfully identifies.

Both matter, but the UPC timetable makes recall particularly important. A narrowly designed search may return a concise set of relevant references and achieve high precision while still leaving important art undiscovered. The limited response period rarely permits every potentially relevant source to be searched exhaustively. A successful strategy must pursue high recall through broad, deliberate and technically focused coverage.

High recall does not mean reviewing sources indiscriminately. It requires complementary search methods directed towards the most promising areas. AI and keyword searching may provide strong starting points, but they should be complemented by citation, inventor and assignee searching, together with review of relevant patent classifications. Expert manual review of carefully selected classes and subclasses can uncover technically relevant documents even when they use language that differs substantially from the asserted patent.

Terminology changes across time, industries and jurisdictions. An earlier reference may solve the same technical problem but describe it using the vocabulary of another field or an earlier generation of the technology. Important disclosures may also appear in a figure, embodiment, appendix or cited exhibit rather than in an abstract or readily searchable passage. International patent sources may require native-language searching and review because machine-translated abstracts do not always capture the relevant disclosure.

Non-patent literature presents additional challenges. Product manuals, data sheets, source code, archived websites, conference papers, draft standards and other technical materials are fragmented across specialist collections and are not consistently indexed. 

These are not reasons to exclude AI from the process. They are reasons to combine its speed and ability to surface connections with the source knowledge and technical judgement of experienced reviewers. AI can help generate candidates and suggest search directions. Human specialists must assess whether unfamiliar terminology, figures or contextual disclosures map to the disputed claim limitations and communicate their technical significance so counsel can develop the invalidity positions.

A diligent search is also shaped by the exchange of knowledge between counsel and the search team. Claim interpretation, infringement positions and evidential needs develop during a case. At the outset, counsel can identify the limitations likely to drive the dispute and explain the shortcomings of the closest known art. As findings emerge, search specialists can show which limitations have been addressed, identify remaining gaps and adjust source priorities in response to legal feedback.

Creative technical and legal analysis becomes particularly important when no single reference discloses every claimed feature. A primary reference may address the limitations said to distinguish the claim, while secondary materials provide evidence that other features were conventional or formed part of the skilled person’s common general knowledge. Developing a persuasive inventive-step position requires more than collecting those documents: the team must explain why the skilled person would have combined their teachings and how the resulting combination bears on the claim as a whole.

This interaction is especially important under the UPC’s compressed timetable. It helps prevent the search from becoming a one-time request followed by a static report. Instead, the search develops alongside the invalidity case, allowing technical findings and legal analysis to inform one another.

The objective is not to choose between AI and human intelligence. It is to design a transparent, expert-led process in which each is used for what it does best. AI can accelerate the search, but human review and creativity, high-recall source coverage and continuous coordination remain central to identifying and developing the strongest prior art under the UPC’s demanding timetable.