EPO Guidelines 2025: What's Changing?

The EPO has released previews of the updated Guidelines for Examination that will come into force on April 1, 2025. Here at Solve Intelligence, we've analyzed the changes to understand their impact on European Patent Practice. Here are just some of the notable proposed changes.

EPO Guidelines 2025: What's Changing?

Use of AI Tools

Firstly, in the General Part of the Guidelines (point 5), a section regarding the use of AI tools has been added:

‘The parties and their representatives are responsible for the content of their patent applications and submissions to the EPO and for complying with the requirements of the EPC regardless of whether a document has been prepared with the assistance of an artificial intelligence (AI) tool.’

The notion that the attorney retains responsibility for their work, regardless of whether they used an AI tool in the process, reflects the recent epi guidelines 3a and 3b in the ‘Use of Generative AI in the Work of Patent Attorneys’. 

Artificial Intelligence and Machine Learning

The Guidelines in G-II-3.3.1 refer to Artificial intelligence and Machine Learning. Following their expansion last year, these guidelines have been further refined in 2025. Firstly, the definition of algorithms has been generalised via removal of specific references to particular techniques like "classification, clustering, regression and dimensionality reduction."

Secondly, further guidance has been added regarding the patentability of AI and ML based inventions. The new guidelines note that: 

‘if a claim of an invention related to artificial intelligence or machine learning is directed either to a method involving the use of technical means (e.g. a computer) or to a device, its subject-matter has technical character as a whole and is thus not excluded from patentability under Art. 52(2) or (3). In such cases, the computational models and algorithms themselves contribute to the technical character of the invention if they contribute to a technical solution to a technical problem, for example by being applied in a field of technology and/or by being adapted to a specific technical implementation.’

This additional contextual information reflects the tried and tested approach to patentability and inventive step that we have seen for many years with respect to computer-implemented inventions. Case law is still light in this area, but decisions like T1669/21 will illustrate more clearly how the EPO intend to assess AI specifically.

Technical Effect in the Problem-Solution Approach

Several updates in G-VII-5.2 relate to technical effects in the problem-solution approach. The Guidelines now explicitly acknowledge that a technical effect submitted during proceedings can be relied upon for an inventive step argument, with T 116/18 confirming these need not be literally disclosed in the original application. However the two criteria from T 1989/19 must be met cumulatively: effects must be encompassed by the technical teaching and embodied by the originally disclosed invention.

Digital Updates

The Guidelines have also been updated to reflect the EPO's ongoing digital transformation. There are updates to the Guidelines in several places regarding the use of MyEPO Portfolio, the EPO's new digital tool for managing communications and correspondence with professional representatives.

There’s information regarding a shared space functionality (detailed in C-VII-2.6, referencing OJ EPO 2023, A59). The ‘shared-area’ allows applicants and examiners to jointly edit documents during consultations and upload and modify documents before and during consultations.

There’s also information regarding fee waivers for MyEPO Portfolio users (A-X-5.2.7), streamlined refunds (A-X-10.3.3) and digital delivery of translations and non-patent literature via MyEPO Portfolio (B-X-11.2)

For the preview of the changes in full, see here.

AI for patents.

Be 50%+ more productive. Join thousands of legal professionals around the world using Solve’s Patent Copilot™ for drafting, prosecution, invention harvesting, and more.

Related articles

Page White Farrer Adopts Solve Intelligence for AI-Assisted Patent Drafting

Page White Farrer Adopts Solve Intelligence and Solve's AI Templates

Page White Farrer ("PWF"), a specialist intellectual property firm with a longstanding focus on technology-driven innovation, has selected Solve Intelligence to support patent application drafting following an extensive evaluation of available AI tools. The firm has initially deployed the platform within its technology-focused patent teams, supporting work across sectors such as AI, software, telecommunications, electronics, semiconductors and engineering. 

The firm's objective was not to automate legal judgement or replace attorney expertise, but to accelerate the preparation of high-quality first drafts of patent applications. By combining AI-assisted drafting with custom templates tailored to the firm's preferred drafting approaches, Solve Intelligence helps attorneys move from an initial claim set to a reviewable first draft more efficiently while maintaining consistency and technical accuracy. 

For this partnership, PWF's primary focus was on configuring the platform around the firm's own drafting conventions, client preferences and established best practices, ensuring that the technology supports the way its attorneys already work. 

The Honest Risks of Using AI in Patent Prosecution

AI is changing how patent prosecution gets done, and the upside is real. So are the downsides: hallucinated citations, confidentiality exposure, export-control pitfalls, and prosecution-history estoppel. This is an honest account of the risks of AI patent software in prosecution, and what actually reduces each one.

Key takeaways

  • Anyone presenting AI-assisted content to the USPTO must reasonably review it; relying on the tool's accuracy alone is not enough.
  • Courts have already sanctioned attorneys for AI-hallucinated citations, and similar risk applies to office-action responses and PTAB briefs.
  • The USPTO warns that AI tools on foreign servers can raise export-control, foreign-filing-license, and national-security problems.
  • Grounded citations, exposed reasoning, zero data retention, and attorney oversight are what materially reduce AI risk in patent prosecution.

Maiwald Chooses Solve Intelligence for Selected AI-Supported IP Workflows

Solve Intelligence is proud to share that Maiwald, one of Germany’s largest and most progressive intellectual property firms, works with Solve Intelligence on selected AI-supported workflows in its patent and legal practice groups. Following an evaluation period during which selected functionalities were assessed for approved internal use cases, Maiwald and Solve have formed an enterprise partnership for 2026.

The rollout reflects the same careful approach Maiwald brings to its client work. Solve enables customers in Europe to process data on servers located within Europe and holds recognized information-security certifications. In addition, Solve offers contractual commitments designed to protect confidential professional information in line with applicable statutory confidentiality obligations and professional conduct requirements. Against that background, any use of AI-supported tools took place only within approved governance structures and under applicable professional, contractual, confidentiality and data-protection requirements.

That balance of innovation and professional judgment is characteristic of Maiwald. The firm has spent decades setting a standard for IP work in Europe: rigorous prosecution, a formidable contentious practice, and interdisciplinary teams of patent attorneys and attorneys-at-law working side by side. The firm approached AI the same way, methodically, with a clear sense of where it adds leverage and where an experienced attorney’s judgment remains irreplaceable. 

PTAB Case Studies of AI Disclosure Requirements: Part II

This article is the second of a series of Patent Trial & Appeal Board (PTAB) case studies (see Part I), which aims to inform applicants’ approach to satisfying the written description and enablement requirements for patenting artificial intelligence (AI) and machine learning technologies.