(John McCann/M&G)
With the sudden, often bewildering onset of artificial intelligence (AI) over the past year, the Big Tech Agenda and the reasons for its oversized influence in the text of the Copyright Amendment Bill seem to have been inspired by a broader agenda to weaken copyright. The aim, to an extent, is to create a safe haven for machine learning. The expanded fair use provisions in the bill appear to be aimed at achieving exactly this.
During the National Council of Provinces (NCOP) public hearings on the bill during 2023, the department of trade, industry and competition conceded that it failed to consider the effects of AI during the development of the bill. It is now too late in the day to do so. This issue would be stood over for a future legislative round.
But once this genie is out of the bottle, it would not be possible to force it back inside. The new copyright exceptions and expanded fair use provisions would create a safe haven for Big Tech to harvest creative content for machine learning and application in their generative AI systems and apps. Although the output of these systems may not infringe copyright, rights would certainly be infringed at the “ingestion stage” if the content is used for machine learning without consent or compensation.
In the US, where lawsuits have already been filed by aggrieved copyright owners, AI developers will seek to rely on fair use and probably also the “2014-Google Books judgment” to defend themselves against claims of copyright infringement. Google Books could be seen to be the first large-scale unlicensed “data grab”. But the US court’s finding of fair use in that matter would not necessarily apply in the AI-machine learning space — the reason being, in the Google Books case, because there was attribution to authors and publishers and only small sections of books were made searchable. The same considerations do not apply to large-scale data grabs that serve no other purpose but to increase the commercial value and offerings of the systems that are trained on the content. (Also, there is no attribution, no permission, and moral rights can also be infringed if the “outputs” distort the original works in certain ways.)
It is in this global context that the department excuse offered to the NCOP select committee that it “ran out of time” to consider AI is astonishing. In the US, the actors and other unions went on strike last year for many months to protect their jobs, careers and employment opportunities for future generations.
So, not only did the department fail to attend to and produce a meaningful economic impact assessment on the bill, it also failed to assess the most pervasive threat to creative content industries and creators’ livelihoods that exist today. The key questions are: how does Big Tech make unlicensed commercial usages of creative content? And how does Big Tech monetise it with no attribution or remuneration flowing back to the original creators, authors and publishers?
One can only wonder whether the creation of a safe haven for machine learning in South Africa was part of the Big Tech agenda pre-2017 when Google co-hosted the infamous workshop with the then department of trade and industry (DTI) and the Freedom of Expression Institute (FXI). It was with regard to this meeting where their influence over the drafting of the “user rights centric” bill was exposed. It may well have been, as the Google Books judgment was handed down in 2014, and the similarities between that unlicensed and large-scale “data grab”, and the current defensive arguments about machine learning (that it does not constitute commercial or expressive usages) are clear to see.
American lawyers at prominent firms have started asking questions on behalf of their clients (AI developers) about when the bill would be enacted. Their clients are already eyeing South Africa as a destination where they could feasibly locate their machine learning operations with a low risk of incurring the same levels of legal liabilities and transparency obligations that they would encounter in the US or European Union. This warning was already offered by a scholarly author who warned that our country could become a safe haven for data piracy. These are, therefore, very real considerations that the department should have had when the opportunity presented itself and when there was still a possibility of conducting a proper socio-economic impact assessment system (SEIAS). A SEIAS should have included considerations of whether the bill would open the door for large-scale exploitative practices in the new era of AI and machine learning. These threaten to disrupt South Africa’s creative industries and further disempower its creatives — moving further away from the objectives of the Creative Industries Masterplan.
Once the bill is enacted, Big Tech and its advocates would hold it up to the World Intellectual Property and the World Trade Organisations as the most progressive legislation for other countries to follow (also members of the African Continental Free Trade Area). A compelling way to frame this legislated data grab is to consider the intellectual labour associated with the sources of the products it seeks to appropriate. Allowing AI companies and technology bigwigs to leech on scholarly publishing (and other intellectual products from the creative economy sector) risks crumpling the creative economy sector of host countries where such resources are currently interwoven with structured knowledge economy infrastructure. It renders futile the in-house publishing rules, creative products, commercial rights, and progressive national policies. Certainly, one may even claim that the human intellectual labourers are being forced to nurture a technology that would eventually make their professional lives difficult while freely benefiting global commercial behemoths inherently unconcerned with the industries they have grounded. (My thanks to Addamms Mututa for the insights in this paragraph.)
This demonstrates how important it is that the creative industries secure a constitutional court ruling that the exceptions and expanded fair use provisions in the bill (and the weak TPM provisions) are unconstitutional and amount to breaches of international law. This would be the only way to achieve a meaningful “reset” in South Africa that would pave the way for the re-drafting of the bill (by recognised copyright experts such as Stellenbosch University professors Owen Dean and Sadulla Karjiker, as well as copyright lawyers Andre Myburgh and Carlo Lavizzari) in a way that is compliant with the South Africa’s Bill of Rights, and its obligations under international law.
Author Profile
Keyan Tomaselli is Distinguished Professor, Dean’s Office, Humanities, University of Johannesburg. This article was first published in ANFASA magazine.