The Battle Over Intellectual Property And Training of Artificial Intelligence Models

The major players in artificial intelligence (AI) development reside in the United States and China. Other nations are wannabes in the race to develop generative AI and artificial general intelligence.

The United States has a history of protecting intellectual property rights. Copyright infringement and fair use of material laws have been enforced by American courts beginning as early as 1790. In 1998, copyright terms extended to 70 years after an author’s death and 95 years for corporate-owned content. Patents are protected for 20 years, and trademark rights since 1946 don’t expire unless abandoned through non-use.

Since joining the World Trade Organization in 2001, China has respected its national intellectual property rights including patents, trademarks, and copyrights within the country and with other nations or groups of nations with which it has signed international agreements. Several cases in China have ruled in favour of foreign company plaintiffs related to intellectual property theft and copyright violations, and abroad, as well Chinese companies have had to pay damages for intellectual and copyright infringements.

One would think that U.S. developers in competition to develop regenerative general-purpose AI, would be seeking protection for their intellectual property and respect that of others, but that appears not to be the case.

AI developers have appealed to Donald Trump and his “America First” agenda with recommendations for an action plan that ignores intellectual property and copyright law based on the argument that the future security of the United States outweighs violations of the ownership of materials being used to train AIs.

Generative AI models need exposure to lots of content to work. While owners of intellectual property used for training are not opposed to AIs using their content as reading material, they are opposed to it being used without paying for it. Some content used by companies like OpenAI, Microsoft, Google and Meta is licensed and paid for. But a significant amount is not and that is leading to lawsuits. News organizations, the film and music industries, and book publishers are seeking court rulings to ensure their intellectual, copyright, and trademark rights are protected.

The American-based AI developers, however, are arguing in court that if they are not given unfettered access to massive amounts of data under fair usage rulings, Chinese AI companies, who supposedly don’t respect intellectual property law will gain an unfair advantage in this new technology race to develop artificial general intelligence.

In the simplistic world of the Trump presidency, American AI companies believe they have found a champion in the competition with China, knowing the president’s “America First” agenda. Trump’s Deputy Director in the Office of Science and Technology Policy has stated, “The Trump Administration is committed to ensuring the United States is the undeniable leader in AI technology.” Does that mean AI companies can run roughshod over intellectual property rights?

Elon Musk, the President’s streamline-government-go-to guy, is also the CEO of xAI, another generative AI American company whose Grok AI is in the race with OpenAI, Microsoft, Google, Meta, and Amazon-back Anthropic with the battle for intellectual property protections on the line.

This February, the U.S. courts rejected a “fair use” defence in a case where AI training used copyrighted legal materials without seeking permission. Will future court rulings extend similar protections to other content such as books, music, artwork and trademarks? It appears that American courts will treat the AI training issue case by case.

What is happening in the U.S. is likely to impact other countries including Canada, the United Kingdom, the European Union and China.

• The 1985 Copyright Act passed by the Canadian Parliament doesn’t reference AI. The evolution of the law since now includes the requirement for AI developers to disclose the datasets used in AI training and encourages licensing agreements if copyrighted material is used. Quebec, whose provincial laws are based on the French Civil Code, protects an author’s moral rights from AI copyright infringement.
• In the United Kingdom, last December the government published changes to the law to make AI developers there happy with an opt-out proposal around data mining of copyrighted material for the training of AI systems. The AI companies receiving the right of use are obliged to disclose what content was being used and to provide training records.
• In the EU, the AI sector faces a bevy of opt-in laws. An across-the-EU copyright law and the EU’s Artificial Intelligence Act require public disclosure of any copyrighted works used for AI training. Companies must obtain licenses and authorization to use these copyrighted materials for AI training.
• In China, meanwhile, the lawful use of copyrighted material for generative AI training requires AI native companies to ensure all data used is legally sourced. Fair use doctrines under Chinese copyright law deny AI companies the right to use online copyrighted content for training without permission from the intellectual property owner.

The American AI companies appealing to Trump state that Chinese AI companies like DeepSeek are a threat to their supremacy and justify their fair use access to copyrighted content as necessary to win the AI race to the top. Chinese copyright law, however, suggests a growing level of compliance with international copyright law.

It, therefore, is hard to understand American AI companies asking the Trump White House to overturn intellectual property protections enforced by acts of Congress dating back to 1790 and 1998. Content developers need fair compensation if their intellectual property is being used by generative AI.