Summary
This edition brings together two stories about AI and intellectual property. The European Patent Office recognised tools that help assess patent portfolios. In the United States, programmers challenged the absence of attribution in AI-generated code, but their claim failed under the particular copyright provision before the Court. Each story below explains the background, the development and its limits.
Reporting period: 11–17 September 2026
Publication date: 18 September 2026
Patents
AI takes a closer look at patent value
A patent portfolio can contain many inventions, but understanding what they are worth requires looking beyond their number. The European Patent Office (EPO) made this the focus of its CodeFest competition, inviting tools to assess the economic and technological value of patents. At this year’s PATLIB conference, the Confused Electrons team won the EUR 20,000 grand prize. Its AI and machine learning platform examines technological impact, market relevance and legal strength to assess a portfolio’s longer-term value.
The other winning entries approached the question differently. Red-Cube coders assessed commercial potential and the extent to which patents might constrain future developments. EUREKA compared inventions with earlier technical knowledge, known as prior art, to assess their inventive contribution. In other words, the competition explored ways of making patent information easier to evaluate. These are competition results concerning valuation tools. The announcement does not change the legal conditions for obtaining a patent or establish that the tools will decide applications at the EPO.
Copyrights
When AI writes code without the credits
The programmers in Doe v. GitHub had shared their code under open-source licences that permitted reuse subject to conditions, including attribution. GitHub Copilot, an AI coding assistant trained on publicly available code, could produce code in response to users’ prompts. The programmers alleged that it reproduced their work without the accompanying credits and licence information. They relied on section 1202(b) of the US Digital Millennium Copyright Act, which protects copyright management information, such as an author’s name or copyright notice, against unauthorised removal or alteration.
The Ninth Circuit upheld dismissal of that claim. As per the Court, the alleged process generated new code without the information; it did not remove information from an existing copy. The Court also clarified that changing a copied work slightly does not necessarily defeat a removal claim. It declined to consider the separate argument about information removed during training because that argument had not been properly preserved. The decision therefore addresses the particular attribution claim before the Court; it does not settle whether training or generated code infringes copyright.
Disclaimer
This post provides general information on developments relating to artificial intelligence and intellectual property. It does not constitute legal advice or a legal opinion. The summaries reflect the cited sources as available on the date of publication. Readers should consult the original sources and seek appropriate professional advice before acting on this information.
References
1. Eur. Pat. Off., CodeFest 2026 on Patent and IP Portfolio (e)valuation, “Winners” (last visited Sept. 18, 2026) (announcing results on Sept. 16, 2026).
2. Doe v. GitHub, Inc., No. 24-7700, slip op. at 10–18 (9th Cir. Sept. 16, 2026).
