OpenAI calls on the Trump administration to grant AI companies an exception to train their models on copyrighted materials. This week, the company published a blog post that was highlighted by The Verge, responding to President Trump’s Action Plan for Artificial Intelligence. Announced at the end of February, the initiative envisages that the White House has asked the private sector to work together to adopt policies that will “strengthen America’s position as a center of power in AI” and promote innovation in the sector.
“America’s robust, balanced intellectual property system has long been key to our global leadership in innovation. We propose a copyright strategy that will expand the system’s role in the intelligence age, protecting the rights and interests of content creators, while also safeguarding America’s leadership in AI and national security,” OpenAI writes in its submission. “The federal government can ensure the freedom of Americans to learn from AI and avoid losing our AI leadership to [the People’s Republic of China] by preserving the ability of American AI models to learn from copyrighted material.”
In the same document, the company recommends that the United States maintain strict controls over the export of AI chips to China. The document also states that the US government should widely implement artificial intelligence tools. By the way, OpenAI started offering a version of ChatGPT intended for use by the US government earlier this year.
This week, Google also published its own list of recommendations for the President’s AI Action Plan. Like OpenAI, the search giant believes that it should be able to train AI models on copyrighted materials.
“Balanced copyright rules, such as fair use and text and data mining exceptions, are critical to enabling AI systems to learn from prior knowledge and publicly available data, unlocking scientific and social progress,” Google writes. “These exceptions allow us to use copyrighted publicly available material to train AI without significantly impacting copyright holders and avoid the often highly unpredictable, unbalanced, and lengthy negotiations with data owners when developing models or scientific experiments.”
Last year, OpenAI stated that “it is impossible to train today’s leading AI models without using copyrighted materials.” Currently, the company is facing numerous lawsuits accusing it of copyright infringement, including those involving The New York Times and a group of authors led by George R.R. Martin and Jonathan Franzen. At the same time, the company has recently accused Chinese artificial intelligence startups of trying to copy its technology.
