Thursday, July 9, 2020
13.00 – 13.10  Opening day 3
Session 3: Data: Copyright in Training Data, Further Rights in Data and Trade Secrets
Issue 8: Copyright Infringement and Exceptions

I speak on behalf of a number of copyright academics that recently released a statement on Implementing User Rights for Research in the Field of Artificial Intelligence: A Call for Action at International Level. That statement was published in short form April 21, 2020, in the Kluwer Copyright Blog, and in long form in the research paper series at www.PIJIP.org https://digitalcommons.wcl.american.edu/research/48.  

As we discuss in that statement, many of the most useful AI projects – including the Blue Dot project that discovered the coronavirus, and hate speech detectors for social media, involve the technical reproduction of copyright-protected works.

Many countries have rights to reproduce materials for research purposes that are broad enough to permit text and data mining to train AI and for other purposes. But many laws are inadequate because they are restricted to non-commercial uses, excerpts of works, or do not extend to the communications between researchers necessary to enable collaboration and validation.

A map of the world coded on the degree to which existing copyright user rights permit TDM research for AI is included below, based on research we will be publishing in the coming months.

WIPO could play a key role in this area.

First, it could serve as an information clearinghouse on legal provisions around the world and the degree to which they allow the reproduction and sharing needed for TDM research.

Second, it could provide advice to countries that are undertaking law reform.

Third, this committee could partner with the SCCR to explore the utility of a normative instrument to facilitate cross border TDM research, such as one based on the Marrakesh Treaty’s novel cross border provision.

Thank you to the committee for its work and I would be happy to consult on these issues further.

Sean Flynn, Andres Izquierdo, Mike Palmedo, Comparing Research Exceptions for Text and Data Mining, PIJIP (2020)