AI: Regulating For Rights
An AI Bill: Where are we now?
In the 2024 King’s Speech, the Government set out its intention ‘to establish the appropriate legislation to place requirements on those working to develop the most powerful artificial intelligence models’ (1). From this announcement, it looked like the Government would be pursuing work to develop an AI Bill in the face of accelerating technological development.
However, in this year’s King’s Speech, there was no AI Bill in sight. The Government announced instead a ‘Regulating for Growth Bill’ (2). Rather than regulating to prevent the human rights abuses that inevitably accompany the endless pursuit of growth, the Government has set out instead that regulation should encourage growth. Regulators will have a strengthened ‘growth duty’ to ‘ensure regulation keeps pace with the unprecedented speed of technological innovation… [and] allow the UK to safely seize opportunities [such as] artificial intelligence (AI)’ (Ibid).
Recently, the AI Bill question found its way back to the Lords again, with a debate on the need for cross-sector AI legislation (3). Led by Lord Holmes, the debate built on his attempt last year to introduce a Private Members’ Bill which called for an ‘AI Authority’ to oversee AI regulation and deliver greater transparency, safety, fairness, non-discrimination and redress (4).
Peers in the debate overwhelmingly supported the Government doing something differently - making recommendations like the introduction of regulation at a national level; a cross-party, pre-legislative scrutiny committee; or the initiation of a Government commission. In the debate, the Government maintained that ‘most AI systems should be regulated at the point of use’ and that the best regulatory approach is the current one (5). The current approach can be considered a patchwork, predominantly made up of fragmented and dispersed ‘technology-neutral’ legislation (like the GDPR’s rules on automated decision making), as well as a number of AI-specific instruments (such as new legislation to govern driverless cars, or the prohibition on AI nudification apps which have been introduced as amendments to existing legislation), all falling under the remit of current regulators.
The problem with regulating ‘at point of use,’ is that it absolves developers from meeting responsibilities to protecting human rights - of their workers, of those who use their products, and those who will be affected by their products. It is at the point of development (if not before) that exploitation begins: and that racialised disparities and the possibility for discrimination become baked in. It also means that systemic issues in how AI systems are built get overlooked, as policymakers focus on narrower downstream issues that arise from specific uses of AI.
When it comes to the big AI developers, their influence on AI policy is outsized. As Chi Onwurah has noted, ‘the reality [is] that these firms have turnovers larger than the GDP of many countries, and their ability to influence stands in stark contrast to that of their users, our constituents, or those campaigning to make the internet safer’ (6).
And indeed, the Regulating for Growth agenda makes explicit that the AI debate is at its heart about capital and power. The Government’s reluctance to deliver a comprehensive, holistic AI Bill is not just a legislative oversight. It is the inevitable end product of capitalist logics which justify the exploitation of people and land to build systems which codify, replicate and extend that colonial rule. As Lelia Marie Hampton has argued, ‘ultimately, the machine learning economy is only made possible through racial capitalism, particularly the exploitation of and pillaging from oppressed racial groups’ (7).
To believe that the AI revolution, as it is currently progressing, will benefit Black women and Black gender-expansive people would mean subscribing to some version of the trickle-down fallacy, where trillionaires being able to further expand their wealth will eventually result in some of that wealth and power returning to racialised groups disproportionately affected by dispossession and poverty. It would mean overlooking how AI is built from the underpaid and insecure labour of people in the Global South (8).
What’s next?
It is increasingly clear that UK AI policy, if it continues on the same trajectory, will not protect people’s rights from discrimination or other violations caused by the rapid development, expansion and mass adoption of AI technologies. This is despite research showing that the public want stronger Government action (9).
This is a global problem. Even in the EU - far ahead of the UK in terms of willingness to bring in oversight of AI that is linked to fundamental rights, with the introduction of the EU AI Act - has since been rolling back their proposals and protections, in the midst of geopolitical anxiety about ceding tech advances - and so, soft and hard power - to the US or China (10).
A next step in the UK AI regulation debate: the Joint Committee on Human Rights will be publishing the findings of its inquiry into Human Rights and the Regulation of AI (you can read the transcript of our oral evidence here).
We hope that, in recognition of its name, the Committee puts the protection of human rights at the very core of its recommendations - and that the Government chooses to place the concerns of the UK public, and not tech oligarchs, at the forefront of their decision-making.