EPO to Use AI Tools for Minuting Oral Proceedings

The European Patent Office (EPO) has announced a pilot project to implement artificial intelligence tools to assist in the preparation of minutes during oral proceedings conducted by videoconference. This development represents a notable evolution in the EPO's approach to procedural documentation, but aligns with the recent trend of increasing AI adoption by the EPO generally.

EPO to Use AI Tools for Minuting Oral Proceedings

Background: Manually-Recorded Minutes

The EPO has consistently maintained that oral proceedings are not to be recorded, by the EPO, parties to proceedings, or the public, with the minutes prepared by the responsible EPO division serving as the only official record. The 2025 Guidelines for Examination (E-III, 10.1) specify that "Sound recordings are made only in the case of taking of evidence. The recording is kept until the end of any possible proceedings. Copies of the recording will not be provided to the parties." This limited exception has been the only context in which any form of recording was permitted during EPO proceedings.

The New AI Initiative

In an announcement dated April 8, 2025, the EPO stated it will begin using AI to assist in preparing minutes for oral proceedings conducted by videoconference. The EPO has provided some clarity regarding how use of an AI tool to this end will be implemented:

  • Oral proceedings held by videoconference before the Receiving Section, examining divisions, opposition divisions and the Legal Division may be audio recorded by the EPO
  • The pilot will initially be limited to proceedings conducted in English.
  • These recordings will be deleted once the minutes of oral proceedings are issued to the parties to proceedings.
  • This will apply from 1 May 2025 
  • The recording or transmittal of oral proceedings in any form by participants other than EPO employees remains prohibited.

Implications for Patent Practitioners

This development reflects the EPO’s growing recognition that carefully implemented and appropriately supervised AI tools may help to enhance procedural efficiency without compromising legal certainty.

For patent attorneys regularly appearing before the EPO, this pilot also raises interesting questions about the future role of AI in proceedings. Although recording proceedings by other participants is still expressly prohibited, if the EPO itself is now utilizing AI tools during oral proceedings, might similar technologies eventually be permitted for use by participating parties?

At Solve Intelligence, we are closely monitoring these developments and the wave of AI adoption across patent offices around the world more generally. The EPO themselves have stated that "The EPO is committed to using AI and machine learning technologies to increase quality and efficiency in the patent grant process [...] closely following a human-centric approach." We share this human-centric philosophy at Solve Intelligence, where we believe in putting patent attorneys in the driving seat of AI adoption. Our approach focuses on augmenting—not replacing—attorney expertise, ensuring that human judgment remains central while leveraging AI to handle routine tasks and improve efficiency. We recognize that the most effective AI solutions for patent practice are those that complement attorneys' specialized knowledge while giving them complete control over the final work product.

Related articles

Hansson Thyresson Integrates Solve Intelligence into Patent Practice​

Founded in Malmö over three decades ago, Hansson Thyresson has built its reputation on close client relationships paired with deep technical and legal expertise. When the firm decided to integrate AI into its practice, it looked for a platform that could deliver enterprise-grade capability.

They chose Solve Intelligence.

Key takeaways

  • Hansson Thyresson has adopted Solve Intelligence for patent application drafting, bringing enterprise-grade AI into the firm's workflows.
  • The firm pairs Solve Intelligence with its hands-on, attorney-led service model, keeping practitioners in control of every output.
  • For Hansson Thyresson, Solve Intelligence amplifies attorney expertise, improving work quality and focus on strategic work that matters most to clients.

Bringing LexisNexis® Global Patent Litigation Data into Solve Intelligence Workflows

Solve Intelligence has teamed up with LexisNexis® Intellectual Property Solutions to bring global patent litigation insights into Solve Intelligence workflows for patent drafting, prosecution, claim charting, freedom-to-operate assessments, invalidity analysis, and related IP work.

Patent Landscape Analysis: How R&D Teams Identify Whitespace

R&D teams use patent landscape analysis to locate uncharted areas of a technology field, assess freedom-to-operate (FTO), and focus development resources on the areas of highest commercial value and broadest potential claim coverage.

Key takeaways

  • A patent landscape search locates whitespace; a novelty search locates prior art blockers. 
  • Patent landscape outputs require contextual analysis; volume alone does not determine opportunity or density.
  • Solve’s Charts tool scans millions of patents from a single file upload and leverages agentic searching, replacing manual iterative keyword & semantic search.

What IPBC Global 2026 in San Diego Revealed about the Future of IP

At IPBC Global 2026, Solve Intelligence was proud to be a sponsor of the conversations shaping the future of intellectual property. At the conference, we demonstrated how our platform amplifies end-to-end patent workflows, including licensing, litigation, and IP-backed finance.

Key takeaways

  • Generic LLMs cannot perform the structured legal reasoning and claim analysis that patent licensing and litigation require.
  • AI lowers the cost of analysing portfolios, from claim-to-product mappings to validity analyses, accelerating both sides of licensing negotiations.
  • IP-backed finance, insurance, and M&A due diligence are growing in sophistication and volume.
  • The profession faces a capacity gap that generic AI tools are not designed to close.