All the content that’s fit to be paid for

Content creators, data producers and businesses are pushing back against generative AI by erecting legal and technical barriers to machine learning (ML). AI training on large language models, such as ChatGPT, is facing copyright infringement lawsuits. More are in the offing. Also, some of the internet’s most-visited consumer and news content sites – Amazon, New York Times and CNN – have blocked GPTBot, a web crawler launched by ChatGPT developer OpenAI. The list, which includes Airbnb and Ikea, is growing rapidly. Social media platforms are not spared either. Although they have some rights on content, ownership rests with users who generate it. Most information on the internet has copyright protection. So, machines, like humans, will be required to pay for using it.

ML deepens the trench between online content and its distribution. Publishers have been seeking a larger share of online advertising revenue. But platforms such as Google and Facebook have been steadfast in denial, arguing content creators earn by employing their reach. Generative AI adds another layer to the issue by introducing an explicit commercial value to the content it sucks up. Humans do the same thing. But the process is less tangible. AI deployed on select data, residing on a specific platform, to solve a unique problem needs rules on how the commercial benefits ought to be shared.

A data regulatory structure should set up guard rails for orderly development of AI. Gov-led AI safety protocols, as recommended by the likes of Microsoft’s Brad Smith, will have to incorporate the treatment of content ownership. It should not be left to the power balance between content creators and AI developers. Blocked content and unacknowledged copyright infringement does not serve the interest of a technology with tremendous transformative potential. National frameworks must mesh at a global level to avoid content localisation, which can slow down ML. Use of content for AI is a case of market imperfection that needs to be addressed by lawmakers.