NICOLE FRITZ: Navigating copyright law and AI development
SA has some complex issues to consider about the intersection of generative AI and copyright
04 July 2025 - 05:00
byNicole Fritz
Support our award-winning journalism. The Premium package (digital only) is R30 for the first month and thereafter you pay R129 p/m now ad-free for all subscribers.
“I agree”, wrote Elon Musk back in April, responding to Jack Dorsey’s dictat posted on X: “Delete all [intellectual property (IP)] law.”
That shared position may come as some surprise to those who imagined Dorsey, Musk’s predecessor at Twitter/X, to be less rapacious than Musk. And it is surprising that a proposal of such wholesale dispossession meets with Musk’s equanimous accord when you think of his rage at the prospect of handing over 30% of any SA Starlink operations to BEE shareholders.
But why are big tech’s titans gunning for IP law, which protects things such as literary and artistic creations, vesting in creators exclusive rights to their work and prohibiting unauthorised use? In short, it is the prospect of copyright impeding the unrestricted development of AI.
In May the US Copyright Office released an extensive report on copyright and generative AI training. At 110 pages it is methodical and scrupulous in its survey — a model of what policy deliberation and design should entail, and hard to reconcile with a government that now trumpets its president’s mercurial “instincts” as a source of superiority.
The report sets out the way great lakes of copyrighted data is provided as input in training AI models, and that generative AI’s outputs also implicate IP law, whether by being substantially similar to or diluting the market for existing copyrighted work. It concludes that generative AI should generally not get a free pass under US copyright law and that licensing agreements must be pursued — that copyright holders be financially compensated for the use of their work by AI developers.
But later that month the head of that office and author of the report was fired by US President Donald Trump. Of course, she’s a woman and so Trump might simply have been pursuing his mission to end “senseless” diversity, equity & inclusion initiatives. But for Martin Wolf, writing earlier this week in the Financial Times, it was at the instance of the “tech bros who had spent so much on bringing Trump to power”. In their eyes, “the report was a declaration of war: it cast doubt on the viability of the ‘fair use’ defence upon which OpenAI, Meta and other tech companies rely for the unrestricted right to ‘scrape’ online data when training their models.”
SA simply can’t afford to be left behind. Picture: 123RF
It should be noted that big tech companies haven’t only been looking to defeat the prospect of copyright law offering impediment to the development of AI. In their sights is any regulation whatsoever. Though now excised from Trump’s one “big beautiful bill”, they strenuously backed the inclusion of a provision that would have imposed a decade long ban on regulation of AI models by individual states in the US.
But lest you think this is all US-centric, think again. Sam Altman’s OpenAI submission to the US Office of Science & Technology Policy earlier this year characterises the EU’s regulations, which permit copyright holders to withhold permission on mining of their work by AI developers, and the UK’s consideration of a similar arrangement, as hindering “AI innovation”. It then exhorts the US government in the service of “American AI leadership and American economic and national security” to “shape international policy discussions around copyright and AI [so as] to prevent less innovative countries from imposing their legal regimes on American AI firms and slowing our rate of progress”.
In the UK more than 400 prominent creatives, media and business leaders have written in support of Baroness Kidron’s amendments to the UK’s Data (Use & Access) Bill, which would require operators of internet scrapers and general-purpose AI models to comply with domestic copyright law: “We will lose an immense growth opportunity if we give our work away at the behest of a handful of powerful overseas tech companies, and with it our future income, the UK’s position as a creative powerhouse and any hope that the technology of daily life will embody the values and laws of the UK.”
And here in SA? Also in May, the Constitutional Court heard arguments as to the constitutionality of the Copyright Amendment Bill, referred there by the president owing to his concerns for its potential conflict with property rights and SA’s international treaty commitments. But while the court’s attention was directed at the importance of our copyright law protecting news producers from having their content ingested and deployed by digital platforms without any fair remuneration (the Campaign for Free Expression and the SA National Editors Forum made submissions to this effect), no direct submissions were made as to how our copyright law should treat use of copyrighted work by generative AI. That seems a significant lacuna, especially in light of Google being said to have played an outsize role in government’s development of the bill.
Neither “copyright” nor “intellectual property” feature even once in our spare eight-paged National AI Framework of October 2024. Nonetheless, the framework confidently predicts that it “lays the groundwork for SA to emerge as a leader in AI innovation while addressing challenges and opportunities in a holistic and sustainable manner”.
Truth be told, SA has achingly complex considerations it will need to navigate regarding the development of AI and the intersection of generative AI and copyright. It simply can’t afford to be left behind — future economic growth and productivity likely depend on it. Still, the prospect of exponentially increased levels of automation and corresponding job shrinkage should terrify this country’s leaders.
Also, it is obvious that SA should want to see the work of SA creatives and scholars being used to train leading AI models and home-grown versions, not least to offset the biases that are ingested in training of AI and that will be reified without such content. But this goal cannot come at the impoverishment of our creatives, news producers and scholars — whose activities and work are so central to our democracy and our sense of what being South African entails.
Any signs that we’re squaring up to this challenge? Sadly, no.
• Fritz is executive director of the Campaign for Free Expression.
Support our award-winning journalism. The Premium package (digital only) is R30 for the first month and thereafter you pay R129 p/m now ad-free for all subscribers.
NICOLE FRITZ: Navigating copyright law and AI development
SA has some complex issues to consider about the intersection of generative AI and copyright
“I agree”, wrote Elon Musk back in April, responding to Jack Dorsey’s dictat posted on X: “Delete all [intellectual property (IP)] law.”
That shared position may come as some surprise to those who imagined Dorsey, Musk’s predecessor at Twitter/X, to be less rapacious than Musk. And it is surprising that a proposal of such wholesale dispossession meets with Musk’s equanimous accord when you think of his rage at the prospect of handing over 30% of any SA Starlink operations to BEE shareholders.
But why are big tech’s titans gunning for IP law, which protects things such as literary and artistic creations, vesting in creators exclusive rights to their work and prohibiting unauthorised use? In short, it is the prospect of copyright impeding the unrestricted development of AI.
In May the US Copyright Office released an extensive report on copyright and generative AI training. At 110 pages it is methodical and scrupulous in its survey — a model of what policy deliberation and design should entail, and hard to reconcile with a government that now trumpets its president’s mercurial “instincts” as a source of superiority.
The report sets out the way great lakes of copyrighted data is provided as input in training AI models, and that generative AI’s outputs also implicate IP law, whether by being substantially similar to or diluting the market for existing copyrighted work. It concludes that generative AI should generally not get a free pass under US copyright law and that licensing agreements must be pursued — that copyright holders be financially compensated for the use of their work by AI developers.
But later that month the head of that office and author of the report was fired by US President Donald Trump. Of course, she’s a woman and so Trump might simply have been pursuing his mission to end “senseless” diversity, equity & inclusion initiatives. But for Martin Wolf, writing earlier this week in the Financial Times, it was at the instance of the “tech bros who had spent so much on bringing Trump to power”. In their eyes, “the report was a declaration of war: it cast doubt on the viability of the ‘fair use’ defence upon which OpenAI, Meta and other tech companies rely for the unrestricted right to ‘scrape’ online data when training their models.”
It should be noted that big tech companies haven’t only been looking to defeat the prospect of copyright law offering impediment to the development of AI. In their sights is any regulation whatsoever. Though now excised from Trump’s one “big beautiful bill”, they strenuously backed the inclusion of a provision that would have imposed a decade long ban on regulation of AI models by individual states in the US.
But lest you think this is all US-centric, think again. Sam Altman’s OpenAI submission to the US Office of Science & Technology Policy earlier this year characterises the EU’s regulations, which permit copyright holders to withhold permission on mining of their work by AI developers, and the UK’s consideration of a similar arrangement, as hindering “AI innovation”. It then exhorts the US government in the service of “American AI leadership and American economic and national security” to “shape international policy discussions around copyright and AI [so as] to prevent less innovative countries from imposing their legal regimes on American AI firms and slowing our rate of progress”.
In the UK more than 400 prominent creatives, media and business leaders have written in support of Baroness Kidron’s amendments to the UK’s Data (Use & Access) Bill, which would require operators of internet scrapers and general-purpose AI models to comply with domestic copyright law: “We will lose an immense growth opportunity if we give our work away at the behest of a handful of powerful overseas tech companies, and with it our future income, the UK’s position as a creative powerhouse and any hope that the technology of daily life will embody the values and laws of the UK.”
And here in SA? Also in May, the Constitutional Court heard arguments as to the constitutionality of the Copyright Amendment Bill, referred there by the president owing to his concerns for its potential conflict with property rights and SA’s international treaty commitments. But while the court’s attention was directed at the importance of our copyright law protecting news producers from having their content ingested and deployed by digital platforms without any fair remuneration (the Campaign for Free Expression and the SA National Editors Forum made submissions to this effect), no direct submissions were made as to how our copyright law should treat use of copyrighted work by generative AI. That seems a significant lacuna, especially in light of Google being said to have played an outsize role in government’s development of the bill.
Neither “copyright” nor “intellectual property” feature even once in our spare eight-paged National AI Framework of October 2024. Nonetheless, the framework confidently predicts that it “lays the groundwork for SA to emerge as a leader in AI innovation while addressing challenges and opportunities in a holistic and sustainable manner”.
Truth be told, SA has achingly complex considerations it will need to navigate regarding the development of AI and the intersection of generative AI and copyright. It simply can’t afford to be left behind — future economic growth and productivity likely depend on it. Still, the prospect of exponentially increased levels of automation and corresponding job shrinkage should terrify this country’s leaders.
Also, it is obvious that SA should want to see the work of SA creatives and scholars being used to train leading AI models and home-grown versions, not least to offset the biases that are ingested in training of AI and that will be reified without such content. But this goal cannot come at the impoverishment of our creatives, news producers and scholars — whose activities and work are so central to our democracy and our sense of what being South African entails.
Any signs that we’re squaring up to this challenge? Sadly, no.
• Fritz is executive director of the Campaign for Free Expression.
Cannes Lions: New rules, old questions
Cloudflare launches tool that blocks AI bot crawlers
Email start-up Superhuman acquired by Grammarly in AI platform push
International business briefs: Moody’s cuts Afreximbank rating
Forget ownership — leasing cars is the future, says Avis COO
Why chat commerce will surpass e-commerce in South Africa
Would you like to comment on this article?
Sign up (it's quick and free) or sign in now.
Please read our Comment Policy before commenting.
Most Read
Related Articles
Cloudflare launches tool that blocks AI bot crawlers
Cannes Lions: New rules, old questions
Email start-up Superhuman acquired by Grammarly in AI platform push
International business briefs: Moody’s cuts Afreximbank rating
Forget ownership — leasing cars is the future, says Avis COO
Published by Arena Holdings and distributed with the Financial Mail on the last Thursday of every month except December and January.