“Text Mining” as Fair Dealing?

At the end of the last class, we realized that one of the pressing questions related to AI that courts around the world will soon have to answer is whether the equivalent of the Canadian “fair dealing” is met when generative AI uses copyrighted works to “learn”.

That same day, I came across an interview that also addressed this question while reading my local newspaper. A Swiss author was interviewed who used Chat GPT as a tool to write his latest novel.

He mentioned that his publisher now not only states on the first pages of his books that the following work is protected by copyright, but also that its use for “text mining” purposes is prohibited:

“Die Nutzung dieses Werks für Text und Data Mining im Sinne von §44b UrhG behalten wir uns explizit vor.» [we explicitly reserve the right to use this work for text and data mining within the meaning of §44b UrhG]

In Germany, where this publisher is mainly active, texts may not be used to train AI – in other words there is no “fair dealing” -, if a corresponding usage restriction has been attached in machine-readable form; see §44b UrhG [German Copyright Act]:

(1) Text and data mining is the automated analysis of one or more digital or digitised works in order to extract information from them, in particular about patterns, trends and correlations.

(2) Reproductions of legally accessible works for text and data mining are permitted. The reproductions shall be deleted if they are no longer required for text and data mining.

(3) Uses according to subsection 2 sentence 1 are only permitted if the rightholder has not reserved them. A reservation of use for works accessible online is only effective if it is in machine-readable form.

Perhaps a role model for the Canadian act as well?

However, the author interviewed is not really convinced by this regulation: “(…) whether this can really be prevented and authors will be fairly remunerated in future if tech companies use their texts to train AI seems questionable to me. Perhaps it will remain wishful thinking.”

In his opinion, the issue should be approached from an economic point of view in particular. He believes, for example, that an “AI-free” label could be introduced for books, similar to the “organic” label for food, and that these books would end up selling better.

Sources (German; translations by deepl.com):