The UK initiates examination of teaching AI systems on copyrighted material.

• The U.K. government launched a consultation on Tuesday regarding regulations for tech companies to use copyrighted content to train their AI models.
• Some artists and publishers are displeased with how their content is being used by companies like OpenAI and Google to train their AI models without permission.

The U.K. is developing guidelines to control how tech firms employ copyrighted material to educate their AI systems.

The British government launched a consultation on Tuesday to enhance clarity for the creative industries and AI developers regarding the acquisition and utilization of intellectual property by AI companies for training purposes.

Some artists and publishers are displeased with how their content is being scraped without permission by companies like OpenAI and Google to train their large language models.

The technology behind today's generative AI systems, such as OpenAI's ChatGPT, Google's Gemini, and Anthropic's Claude, is based on large language models.

In 2020, The sued and OpenAI for copyright infringement and intellectual property abuse, alleging that they used the companies' data to train large language models.

OpenAI challenged the NYT's accusations, asserting that utilizing open web data for AI training should be considered "fair use" and that it offers an "opt-out" for rights holders "because it's the right thing to do."

In the U.K., Getty Images filed a lawsuit against Stability AI, alleging that it scraped millions of images from its websites without permission to train its Stable Diffusion AI model. Stability AI, however, has contested the lawsuit, stating that the training and development of its model occurred outside the U.K.

Proposals to be considered

The consultation will examine the possibility of granting an exception to copyright law for AI training in commercial settings, while still allowing rights holders to retain control over the use of their content.

The consultation will propose measures to help creators license and be compensated for the use of their content by AI model makers, while also providing clarity to AI developers regarding the material that can be used for training their models.

To ensure effective, accessible, and widely adopted standards and requirements for rights reservation and transparency, both the creative industries and technology firms must do more work, as stated by the government.

The government is considering proposals to increase transparency in AI model training datasets and acquisition methods, allowing rights holders to understand how their content has been used to train AI.

Technology companies are not transparent about the data and methods used to train their algorithms, as it involves commercial secrets that could be exploited by competitors.

Under the leadership of former Prime Minister Rishi Sunak, the government endeavored to establish a voluntary code of practice for AI copyright.

AI copyright rules: U.K. versus U.S.

In a recent interview with CNBC, the head of an app development software company stated that he believes the U.K. is well-positioned to become the world's leading authority on this issue.

"Matt Calkins, Appian's CEO, stated to CNBC that the U.K. has declared its commitment to protecting personal intellectual property rights. He pointed to the Data Protection Act of 2018 as evidence of the U.K.'s strong association with intellectual property rights."

The U.K. may not be as susceptible to pressure from domestic AI leaders as the U.S., according to Calkins, who noted that the country is not "subject to the same overwhelming lobbying blitz."

Before a bill about AI is passed in the U.S., it will be met with opposition from Amazon, Oracle, Microsoft, or Google, according to Calkins.

"The major AI players are taking wholesale intellectual property without protecting the rights of individuals, making it difficult for anyone to write sensible legislation."

As technology companies shift towards a more multimodal approach to AI, the issue of potential copyright infringement by AI firms is gaining more attention.

Last week, OpenAI released its AI video generation model Sora in the U.S. and "most countries internationally." The tool enables users to input a desired scene and generate a high-definition video clip.