The Trump administration seems to be undecided about whether artificial intelligence should freely learn from the work of Americans. In a recent court filing, the Department of Justice (DOJ) asserted that training AI on copyrighted books constitutes fair use, arguing that requiring licenses would primarily benefit large labs and well-established media outlets. This stance contrasts with the White House’s AI framework from March, which suggested that it’s still unclear if such training breaches copyright laws.
The DOJ emphasized that allowing tech giants to dominate large language model (LLM) training would not serve the public interest. It stated, “The United States has a strong interest in this Court rejecting any argument that training LLMs on copyrighted texts violates copyright law.” The department contended that OpenAI uses these works to educate its models rather than to reproduce them. According to the DOJ, “An OpenAI LLM uses the copyrighted work not to duplicate the work’s expressive content, but as part of a process to learn and act on statistical patterns.”
When asked for a comment, the DOJ referred to a September 2 post by associate attorney general Stanley E. Woodward Jr. He highlighted President Trump’s view that maintaining AI superiority is essential for national security and economic advancement. Woodward reiterated that the administration would not allow misunderstandings of copyright law to put the nation at a disadvantage compared to foreign competitors.
Interestingly, one tech policy expert, Evan Swarztrauber, did not agree with the DOJ’s position. He pointed out that while the administration acknowledges the complexities of litigation surrounding AI training and copyright, it also noted that ultimately, it’s up to the courts to decide. Swarztrauber critiqued the DOJ for its assertive stance that nearly all AI training qualifies as fair use and described its position on licensing as overly negative. He stressed that licensing should not be dismissed as just a means to support established media companies, as smaller creators struggle to protect their rights and deserve compensation.
In 2023, The New York Times filed a lawsuit against OpenAI, claiming that its ChatGPT models were trained using numerous articles from the outlet. The Times argued that these AI models were competing against its journalism by generating outputs that can replicate or closely summarize its content. Microsoft was also implicated in this lawsuit, with allegations that its Copilot featured verbatim sections of Times articles in chatbot responses. Similarly, the company Anthropic faced accusations for building AI models using copyrighted books. It was discovered that its Project Panama involved purchasing and then destroying the spines of millions of books to scan their pages for model training. A video circulating on social media showed this process, igniting further concern over the treatment of literary works in the AI landscape.
