The New York Times recently filed a lawsuit against OpenAI, which operates ChatGPT, and Microsoft, which collaborates with OpenAI on the platform Copilot, over the defendants’ use of the newspaper’s copyrighted articles.

In its lawsuit, the New York Times alleged that OpenAI and Microsoft used millions of copyrighted articles, without its permission, to train their AI platforms, all while infringing on the articles’ copyrights. The newspaper is demanding that the companies destroy not only the database of articles they operated, but also all of the models they created through use of those articles. The New York Times is also seeking compensation in the volume of billions of dollars.

School of Algorithms

AI-based generative models, such as ChatGPT, actually “learn” and “train” using immense databases and information bases. This enables them to generate the desired output for the user via image, text, etc. Without these databases, the models would be empty of content and meaningless. Only the use of these raw databases impart AI-based models with their phenomenal capabilities. The more numerous and diverse the databases, the higher the quality, resolution, and accuracy of the output generated.

However, this methodology could adversely impact entities, perhaps even significantly, that own digital media on a massive scale, such as the New York Times. Therefore, the newspaper’s view is that, if it has no entitlement to legal protection for its products, it will stop producing them, leaving a void that no computer or AI platform can fill.

Who Owns the Rights?

This conflict is another fascinating link in the series of conflicts between database owners and entities operating AI platforms. For example, a few months ago, Getty Images filed a lawsuit against the Stability AI platform, both in the United States and in England, alleging copyright infringement due to unauthorized use of its image databases. A group of visual artists also filed a lawsuit against both Stability AI and the Midjourney platform. The platform operators’ main argument within this context relies on “fair use” protection for the purpose of using the information and training the platforms, while the database owners are claiming infringement of their copyrights and an adverse impact on creative incentives, insofar as the AI platforms made use of their works without their consent.

In the local sector, the Israeli Ministry of Justice published its position that the use of copyrighted content within the machine-learning context is, in most cases, permissible under the existing Copyright Law, and therefore does not constitute copyright infringement. This is because such use will usually be fair use, and because the protections under the provisions of incidental use and transient use may apply. This position is in line with Israel’s aspiration not to be “left behind” in the race to develop AI systems and to incentivize desirable and competitive activities in this evolving sector. We emphasize that Israeli courts have not yet examined this position, and we expect that it will take time for this to happen.

Looking Ahead

A United States court ruling on this issue could dramatically change the framework of the relations and obligations between database owners and companies developing AI tools. Looking ahead, it is likely we will see the promotion of business and legal solutions to mitigate and manage AI developers’ risks involved in using other companies’ databases and media. One of the solutions may be mapping the key copyright holders, and AI developers applying to them for a license to use to teach and train their AI platforms. Another possible solution is a “coercive license,” which will permit the use of the copyrighted products, subject to royalties at a reasonable rate. A third possible solution may be to activate an “opt-out” mechanism with respect to copyright holders, under which any party that does not want a creation belonging to it included in the database will be able to demand its removal from the database (similar to the “opt out and data removal” mechanism on the YouTube portal).

Companies that use and train AI tools must ensure they use information transparently, while verifying and checking the databases used for machine learning. Use of copyrighted materials should be done carefully, with an understanding of the legal risks and potential repercussions.

[View source.]