Exclusive:Trump administration enters AI legal row with Encyclopaedia Britannica
US government says training of AI models crucial for national security reasons
The US government has urged a court to throw out a legal case brought by an Edinburgh founded firm against one of the world’s leading artificial intelligence companies, arguing that its use of centuries’ worth of research does not amount to a breach of copyright law.
Encyclopaedia Britannica, which is synonymous with the Scottish Enlightenment, is suing the AI giant, OpenAI, in order to protect its meticulously researched content.
As revealed by The Scotsman earlier this year, the company, founded in Edinburgh in 1768, has accused one of the tech world’s most prominent entities of “cannibalising” the work produced by generations of researchers, writers and editors for AI summaries which, it says, flout copyright and trademark legislation.
The legal action, brought in the US, accused OpenAI’s ChatGPT products of taking a “free ride” on Encyclopaedia Britannica’s “high quality” and “trusted” content, meaning it is being denied crucial advertising and subscription revenues.
The company, which published the first three volume edition of its renowned encyclopaedia in 1771, is one of several taking legal action over OpenAI’s training of large language models (LLMs), with other plaintiffs including the New York Times.
However, OpenAI has rejected the accusations brought by Encyclopedia Britannica, arguing that its models “empower information,” and are trained using “publicly available data.”
Now, senior officials at the US Department of Justice (DoJ) have waded into the dispute, arguing that any attempt to hamper OpenAI’s LLM training would impact US national security.
Its submission to the US District Court in New York argues that the US has a “strong interest” in the question of whether training AI models on written works constitutes ‘fair use’ under copyright law, noting that effective integration of AI is crucial for its US security capabilities, with the technology needed to enhance weapon systems, analyse intelligence, and provide information on where best to target missile strikes.
The submission, made by Michael Weisbuch, deputy associate attorney general at the DoJ, states: “Rules of law that make it significantly more difficult to develop a robust AI industry in the United States therefore threaten national security and give a competitive advantage to foreign adversaries who are not so encumbered.”
It adds that given the national interest in promoting innovation and free expression, the training of AI models on copyrighted material did not violate copyright law, and argues that at a minimum, any effect on the market for copyrighted works “cannot overcome the significant, world-changing value” of the “transformative” purpose of training AI.
The filing also points out that as recently as June, President Donald Trump issued an executive order reiterating the nation’s commitment to encouraging AI innovation and “accelerating responsible AI adoption” across government and industry.
The complaint, filed by Britannica in conjunction with Merriam Webster, which it owns, claims OpenAI has used nearly 100,000 of Britannica's online articles for use as "training inputs" for its AI language models, but stressed that the "true extent" of how much content is copied was only known to the the tech behemoth.
The case continues.
Comments
Want to join the conversation? Please or to comment on this article.