Written by Professor Yuval Shany, Senior Inaugural Fellow of the Accelerator Fellowship Programme, Institute for Ethics in AI and Dr Caroline Emmer De Albuquerque Green, Director of Research, Institute for Ethics in AI
Warnings by tech leaders to slow down AI development to avert a threat to human existence; a dramatic incident report by the frontier AI lab ‘Anthropic’, describing attempts to use its large language model, ‘Claude’, to generate swarms of AI agents that would conduct widespread cyberattacks, play a role in criminal operations and even the development of conventional and unconventional weapon systems. These events unfolded within a single week – nothing particularly unusual in the world of AI news.
Indeed, the threats to human dignity and fundamental human rights posed by AI are becoming clearer, although new risk scenarios keep emerging. While the development of regulatory mechanisms and other responses to AI-related risks has not been able to keep up with the pace of AI development, the UK Parliament’s report on “human rights and the regulation of AI”, published by the Joint Committee on Human Rights on 14 September 2026, is a step in the right direction towards addressing the glaring regulatory gap in the United Kingdom.
Based on the Committee’s analysis of the adequacy of existing UK laws and institutions to effectively protect human rights from AI-related threats, the report concludes that there is a need for AI-specific legal reform, steering away from the current sector-driven approach to responding to AI through existing legal frameworks towards one that applies to all sectors. It also calls for an institutional reform, specifically the creation of a dedicated AI oversight body.
Although limited by its national focus, the report provides important impetus by highlighting human rights-related harms, regulatory gaps and the need for new pathways to address risks and violations in the context of AI systems that are developed and controlled by private companies. More generally, the report endorses three important propositions: 1) There is no contradiction between AI innovation and regulation, and a human rights-centred regulatory scheme may actually help the UK maintain a position of global leadership in the transition to an AI economy. At the same time, any regulation must be proportionate – i.e., not overly burdensome on businesses and government; 2) The societal risk from AI is rapidly increasing (noting the Hugging Face incident – involving the autonomous release of cyber-attacking swarms of AI agents – as illustrative of the new threat environment confronting critical infrastructure), requiring a robust regulatory response, including the prohibition of certain systems (following public consultation) or prior approval before marketing or deployment of high-risk AI systems; and 3) UK regulation should be based on international standards. In this regard, the report calls on the UK to ratify the Council of Europe AI Convention.
The Joint Committee on Human Rights’ report and its suggestions
The Joint Committee report analyses the effects of the deployment of AI systems on human rights, using as a point of departure the Council of Europe 2024 Framework Convention on Artificial Intelligence and Human Rights, Democracy and the Rule of Law. Alongside traditional concerns about privacy and discrimination, it considers evidence about the implications of opacity for accountability (the right to an effective remedy), the use of surveillance to exploit the vulnerabilities of online users, and the problems associated with increased resort to automated decision-making. It then identifies a number of normative and institutional shortcomings in UK law and proposes legal reform.
In the absence of comprehensive AI legislation, the protection of human rights in this field is currently undertaken by different legal instruments, most of which are not AI-specific. This approach leaves in place significant protection gaps: for example, transparency requirements apply mostly to public rather than private uses, responsibility attaches to downstream operators (at the point of deployment) – who are often ill-placed to prevent harm – and not to upstream developers who are better situated to address human rights risks, and the right to a human decision is often illusory in practice due to the limitations of human supervision. On the institutional level, the report finds that the current regulatory landscape in the UK is fragmented, sectoral and harm-driven. There is no single central AI regulator that can engage with technology developers, quickly produce needed standards and exercise powers of monitoring and enforcement, including conducting pre-market evaluations. It also notes barriers to the effective enforcement of rights, including the speculative, uncertain and costly nature of litigation.
As a result, the report concludes that comprehensive legal reform is needed.
First, it recommends the passage of a new AI act that would embed human rights protections. The envisioned new legal framework would comprise a system of risk assessments and differentiated duties, including due diligence obligations, attached to different relevant actors in the AI supply chain (upstream and downstream). Whereas the new law should set out generally applicable principles, detailed guidance should be introduced through periodically updated “codes of practice”. As for substantive standards, it is recommended that certain uses of AI systems be banned altogether (for example, “subliminal techniques, emotional inferences or inappropriate use of profiling or biometric data”); for other systems, prior approval would be required. In assigning different regulatory consequences to different risk levels, the report recommends following the general approach of the EU AI Act of regulating through a system of risk assessments and mitigation measures. The report further recommends a comprehensive transparency requirement, including a degree of explainability, and the enactment of more robust protections against the misuse of automated decision-making.
In addition, the report suggests the creation of a single AI regulatory body in the UK.
Such a regulatory body may be built on existing ones (e.g., the AI Security Institute), but would require, in that case, a significant expansion in powers, responsibilities and capacities. The report recommends that the new body be responsible inter alia for issuing codes of practice, evaluating AI systems, collecting incident reports, investigating breaches, sanctioning violators and prohibiting the use of unacceptably risky AI systems. We could not agree more with the recommendation to establish a single regulator, as such a dedicated regulatory body would focus on the technology and its implications directly, rather than the current approach in which regulatory bodies consider AI as one aspect of their work, often without the necessary expertise in either AI or human rights.
The need to provide more specific international guidance
With regard to the substantive alignment of UK regulation with human rights standards, we note that the Council of Europe AI Convention, which the report favourably cites, uses language that is often too general in nature, and offers only a rudimentary list of the main human rights particularly affected by AI systems. A White Paper published by one of us last year under the auspices of the Institute for Ethics in AI seeks to promote a multi-stakeholder consensus around more detailed normative standards that would go beyond the language of the Council of Europe AI Convention. An instrument elaborating such normative standards would better guide the application of existing human rights to different uses of AI systems and foster the gradual emergence of new AI human rights.
The list of rights enumerated (and elaborated upon) in the White Paper partly overlaps with the rights discussed in the Joint Committee report. It includes the following rights:
- The right of access to AI systems
- The right to privacy-related protections from harmful uses of AI systems
- The right to be free from algorithmic bias and unfairness
- The right to algorithmic transparency and explainability
- The right not to be subject to algorithmic manipulation
- The right to a human decision and a human-to-human interaction
- The right to accountability for harms caused by the use of AI systems
Like the Joint Committee report, the White Paper underscores the need to embed such AI human rights in the applicable normative framework – at this point in time, most probably in an international soft-law declaration or guidance document. The proposed UK regulation could lend important support and momentum to such an international standard-setting initiative. In due course, it could itself be informed by further emerging international standards.
The Oxford AI and Human Rights Summit
The interplay between national, regional and international standard-setting and legal reforms is at the heart of the Oxford AI and Human Rights Summit, which will take place on 5 October 2026. The Summit will engage human rights experts from national, regional and global systems, national regulators, commissioners and judges, and civil society and AI industry representatives in a conversation about existing protection gaps, needed normative and institutional reforms and practical steps to advance a reform agenda. The Joint Committee report offers Summit participants a good template for how to address such questions in a particular national context, which takes seriously the challenge of adapting an existing human rights framework to the opportunities and risks associated with the use of AI systems. In this regard too, it is a step in the right direction.
-
Suggested citation: Professor Yuval Shany and Dr Caroline Emmer De Albuquerque Green, ‘A step in the right direction: The Joint Committee on Human Rights calls for human rights-friendly AI regulation in the UK’, (29 September 2026) AI Ethics at Oxford Blog; https://www.oxford-aiethics.ox.ac.uk/blog/step-right-direction-joint-committee-human-rights-calls-human-rights-friendly-ai-regulation-uk