Several Canadian news media holdings are suing American company OpenAI, alleging that OpenAI scraped large quantities of information from Canadian sources and used that data to train, among other things, ChatGPT. This, according to the plaintiffs, was done without authorization.
The plaintiffs seek both an injunction against further data scraping, as well as an unspecified amount in damages from OpenAI.
It will be interesting to see how Canadian courts interpret this under fair dealing. OpenAI may argue that their use falls under either the discrete research or education fair dealing categories. In turn, the Canadian companies might argue the character, amount, and effect of the dealing were unfair.
Other jurisdictions seem to be wrestling with the issue. Our colleague Alex posted about the German/Swiss perspective https://iplaw.allard.ubc.ca/2024/11/05/text-mining-as-fair-dealing/. US courts seem to have allowed large-scale data scraping, given how a similar lawsuit brought forth by some US media companies was tossed out. It remains to be seem how this latest development will play out in Canada.
Source: https://www.reuters.com/sustainability/boards-policy-regulation/major-canadian-news-media-companies-launch-legal-action-against-openai-2024-11-29/
Copyright & Social Media
Communications Law
This is an interesting post. I think one key argument that OpenAI can make is that most of their users do not pay a fee, and among the small percentage of paying customers, it is questionable how scrapped information from these news outlets is used for profit. Admittedly, some people possibly rely on AI-generated news to produce their own work, arguably committing secondary infringement. However, even though this is the case, they are likely just “pirating” the facts that are not copyrightable. On the other hand, if there is evidence that OpenAI has scrapped news articles from subscriber-only content, such conduct may affect the revenue of those news outlets as ChatGPT users can bypass subscription fees.
This is an interesting post. I think one strong argument OpenAI can make is that it doesn’t profit from providing summaries of news (which ChatGPT does) to its users, considering that most Chat users don’t pay a fee. OpenAI can also question how many paying users will use scrapped news for commercial purposes. Even if some people build their own work based on “pirated” news, they are most likely just using the facts, which are not copyrightable. On the other hand, if there is evidence that Chat gave out subscriber-only articles, those news outlets may claim revenue loss because users can bypass subscription fees. This may undermine its fair dealing defence.