7BR Chambers

Asuccessful first Global Dialogue on AI Governance should move beyond high-level principles to deliver practical, globally relevant foundations for action. First, it should establish a clear shift from reactive regulation to preventative design, recognising that many AI-related harms originate at the point systems are created. Embedding Equality by Design ensuring fairness, inclusion and non-discrimination are considered at the outset should be identified as a core governance objective. Second, the Dialogue should produce shared baseline standards across jurisdictions, particularly in relation to: I. ex ante risk and equality impact assessments II. transparency and explainability of AI systems III. meaningful rights to challenge automated decisions Without a degree of interoperability, fragmented approaches risk undermining both trust and effectiveness. Third, success would include the creation of practical mechanisms for accountability, including independent oversight, auditability of high-risk systems, and accessible routes to redress. Principles alone are insufficient without enforcement. Fourth, the Dialogue should ensure meaningful inclusion of diverse and underrepresented voices, particularly those most affected by AI systems. This includes civil society, workers, and communities who experience the real-world consequences of automated decision-making. Finally, a successful outcome would position the UN as a coordinating and standard-setting body, capable of bridging legal, technical and policy perspectives, and supporting ongoing global cooperation. At its core, success will be measured by whether this Dialogue begins to translate shared concern into coherent, actionable and equitable governance frameworks.

• Safe, secure and trustworthy AI
• Social, economic, ethical, cultural, linguistic and technical implications of AI
• Protection and promotion of human rights
• Transparency, accountability, and human oversight

My selected priorities reflect a core concern: that AI systems risk replicating and amplifying existing inequalities unless fairness is embedded at the point of design. "Safe, secure and trustworthy AI" cannot be achieved through technical robustness alone. Trust is fundamentally linked to whether systems operate fairly, transparently, and without discrimination. This is why the protection and promotion of human rights, alongside transparency, accountability and human oversight, are essential governance pillars rather than complementary considerations. In my work as a barrister in employment and human rights law, I have seen how ostensibly neutral systems can produce discriminatory outcomes when underlying assumptions go untested. As AI systems increasingly shape access to work, services and opportunity, these risks are magnified at scale. For this reason, I advocate an "Equality by Design" approach. This framework emphasises: Embedding fairness and inclusion at the outset of system development Ensuring meaningful transparency and the ability to challenge decisions Establishing clear accountability and routes to redress The inclusion of the social, economic and ethical implications of AI is therefore critical. AI governance cannot be reduced to technical standards alone; it must engage with the real-world impact of systems on individuals and communities, particularly those most vulnerable to exclusion. Taken together, these priorities reflect the need for a holistic approach to AI governance one that integrates legal, technical and societal perspectives and moves from reactive regulation to preventative, system-level fairness.

Key cross-cutting issue not yet fully captured is the need to shift from reactive governance to preventative, design-led approaches. Current AI governance frameworks largely focus on managing risk after systems are deployed. However, many harms particularly those relating to discrimination and exclusion, originate at the point of system design. Without addressing this upstream stage, governance risks becoming remedial rather than transformative. This points to the need for a stronger emphasis on design accountability: Who is responsible for decisions made during system development How assumptions, data selection and optimisation goals are interrogated How potential disproportionate impacts are identified and mitigated before deployment A related emerging issue is the scaling of systemic bias through automation. AI systems do not simply replicate individual instances of unfairness; they can entrench patterns across entire sectors, often in ways that are difficult to detect or challenge. This raises questions about collective harm, evidential thresholds, and access to justice-areas where existing legal frameworks are not yet fully equipped. There is also a growing need to consider the interaction between AI systems and existing institutional structures, including workplaces, public services and regulatory bodies. AI does not operate in isolation; it is layered onto systems that may already contain embedded inequalities. Addressing these issues requires closer integration between legal, technical and policy disciplines, and a greater focus on embedding fairness, transparency and accountability at the earliest stages of system development.

In the UK and across comparable jurisdictions, there is a growing recognition of the importance of AI governance, with developments such as emerging regulatory guidance and alignment with international frameworks. However, significant gaps remain. A key challenge is that existing legal frameworks particularly in equality and employment law are largely reactive, addressing harm after it has occurred. AI systems, by contrast, can produce and scale discriminatory outcomes rapidly, often without clear visibility or straightforward routes to challenge. This creates a mismatch between the speed and scale of AI decision-making and the capacity of legal systems to respond effectively. There are also gaps in transparency and accountability. Individuals affected by automated or AI-assisted decisions may struggle to understand how those decisions were made, particularly where systems are complex or proprietary. This raises concerns about access to justice and the practical enforceability of rights. From a sector perspective, I have observed increasing concern around algorithmic decision-making in workplace contexts, including recruitment, performance management and dismissal. These systems risk embedding systemic bias where underlying data or assumptions are not critically assessed. However, there are also clear opportunities. The current moment presents a chance to move towards more preventative, design-led governance approaches, including ex ante risk and equality impact assessments. There is also increasing appetite among policymakers, regulators and organisations to engage with ethical and human rights-based frameworks. Bridging legal, technical and policy expertise will be essential. With the right approach, AI governance can move beyond risk management towards embedding fairness, accountability and trust at the heart of system design.

The AI Dialogue can play a critical role as a global convening and standard-setting platform, bridging fragmented national and regional approaches to AI governance. At present, there is a growing divergence in regulatory models across jurisdictions. While this reflects different legal traditions and policy priorities, it risks creating inconsistency in standards, uncertainty for organisations, and uneven protection for individuals. The Dialogue can help establish shared baseline principles and practical frameworks that promote interoperability while respecting local context. In particular, the Dialogue can advance cooperation by: I. Facilitating alignment on minimum standards for transparency, accountability and human rights protections II. Promoting the adoption of ex ante risk and impact assessment approaches across jurisdictions III. Enabling the exchange of best practice between regulators, policymakers, industry and civil society The UN is uniquely placed to ensure that this cooperation is inclusive and globally representative, incorporating perspectives from both developed and developing economies, as well as those most affected by AI systems. Importantly, the Dialogue can move beyond principle-setting to support practical coordination mechanisms, such as shared methodologies for auditing high-risk systems, and common approaches to oversight and redress. By fostering collaboration across legal, technical and policy domains, the Dialogue can help ensure that AI governance evolves in a way that is coherent, rights-based and capable of addressing cross-border challenges.

The AI Dialogue should build upon and connect with existing international and regional initiatives, including emerging regulatory frameworks, standards-setting bodies, and multi-stakeholder partnerships focused on responsible AI. This includes, for example: I. Regional regulatory approaches such as the EU AI Act II. Global principles developed by organisations such as the OECD and UNESCO III. Cross-sector initiatives involving industry, academia and civil society While these initiatives provide important foundations, they remain fragmented in scope and application, with varying levels of enforceability and alignment. The added value of the AI Dialogue lies in its ability to act as a coordinating and integrative mechanism, bringing these strands together into a more coherent global approach. In particular, the Dialogue can: I. Promote greater interoperability between existing frameworks II. Support the translation of high-level principles into practical, operational standards III. Ensure that human rights and equality considerations are consistently embedded across initiatives A further area of added value is the opportunity to emphasise preventative, design-led approaches to governance. Many existing frameworks focus on risk management and mitigation, but less attention has been given to embedding fairness and accountability at the point of system design. By aligning existing initiatives around shared goals and strengthening their practical application, the AI Dialogue can help move global AI governance from a collection of parallel efforts towards a more coherent, effective and equitable system.

Different stakeholders bring distinct and complementary perspectives to AI governance, and the Dialogue should be structured to reflect this diversity in a meaningful and practical way. Governments and regulators can contribute legal frameworks and policy direction; industry can provide technical expertise and implementation insight; academia can offer research and critical analysis; and civil society and affected communities can bring essential lived experience of how AI systems operate in practice. To be effective, the Dialogue should move beyond traditional panel formats and adopt a multi-layered structure, including: I. Thematic working groups focused on specific governance challenges II. Cross-sector roundtables that bring together legal, technical and societal perspectives III. Mechanisms for capturing and integrating lived experience, particularly from those directly impacted by AI systems There should also be a clear emphasis on continuity and output, with structured pathways from discussion to recommendation and implementation. This could include: a) The development of practical guidance or model frameworks b) Iterative feedback loops between sessions c) Transparent reporting on outcomes and next steps Importantly, the Dialogue should create space for interdisciplinary exchange, recognising that effective AI governance sits at the intersection of law, technology and society. A well-designed structure will ensure that contributions are not only heard but translated into coherent, actionable outcomes. There is also a need to ensure accessibility, including through language, format and timing, to enable meaningful global participation. Embedding these perspectives is essential not only for fairness, but for the effectiveness of AI governance itself. Systems that are designed without diverse input are more likely to produce unintended and inequitable outcomes.

Global discussions on AI governance often underrepresent those most directly affected by AI systems. This includes workers subject to algorithmic management, individuals impacted by automated decision-making in areas such as recruitment, healthcare and public services, and communities that have historically experienced structural inequality. Voices from the Global South are also frequently underrepresented, particularly in shaping governance frameworks that may ultimately be applied across diverse contexts. In addition, there is a need to amplify perspectives from disciplines beyond technology, including law, social sciences and community-based organisations. The consequence is that AI governance risks being shaped by those who design and deploy systems, rather than those who experience their impact. Addressing this requires more than representation in principle; it requires structural inclusion. This could include: I. Targeted outreach and support to enable participation from underrepresented groups II. Dedicated forums for affected communities to share lived experience III. Integration of those perspectives into formal decision-making processes, rather than treating them as peripheral input.

To foster meaningful and dynamic engagement, the AI Dialogue should adopt formats that move beyond passive discussion towards active, problem-solving collaboration. One effective approach would be the use of scenario-based workshops, where participants engage with realistic case studies involving AI systems in practice (for example, in employment, healthcare or public services). This allows stakeholders to explore governance challenges in context and develop practical solutions collaboratively. Another valuable format is interdisciplinary labs, bringing together legal experts, technologists, policymakers and civil society to work through specific issues, such as bias mitigation or transparency requirements. This can help bridge the gap between principle and implementation. The Dialogue could also incorporate evidence sessions, where individuals and communities share lived experiences of AI systems, ensuring that governance discussions remain grounded in real-world impact. Digital participation tools can further support inclusivity and global engagement, enabling contributions from a wider range of stakeholders across different regions. Finally, there should be a focus on output-oriented sessions, where discussions lead to tangible outcomes such as draft frameworks, recommendations or model standards. Innovative engagement should not be an end in itself, but a means of ensuring that the Dialogue produces practical, inclusive and actionable results.

An effective approach to AI governance is to move from reactive regulation towards preventative, design-led frameworks. One such approach is "Equality by Design", which seeks to embed fairness, inclusion and accountability into systems at the point of creation, rather than addressing harm after it occurs. Equality by Design operates across three practical pillars. First, design-stage assessment: this includes ex ante equality and risk impact assessments, rigorous testing for bias across protected characteristics, and critical evaluation of data sources and optimisation criteria. This approach mirrors established practices such as data protection impact assessments, but applies them more explicitly to equality and non-discrimination. Second, meaningful transparency: systems should be explainable and contestable in practice, not only in principle. This includes clear communication of how decisions are made, the limitations of data, and accessible mechanisms for individuals to challenge outcomes. Third, accountability and redress: governance frameworks must include ongoing monitoring of outcomes, independent oversight of high-risk systems, and effective routes to remedy where harm occurs. Existing initiatives, such as emerging regulatory frameworks, international principles, and sector-specific guidance provide important foundations. However, many focus on risk management after deployment. Equality by Design adds value by shifting the focus upstream, ensuring that fairness is built into systems before they are operational. This approach is particularly relevant in high impact contexts such as employment, healthcare and public services, where automated decision-making can significantly affect individuals' rights and opportunities. Embedding Equality by Design within global AI governance would support the development of systems that are not only innovative, but fair, trustworthy and aligned with human rights from the outset. If we design systems with architecture of equality in mind, we reduce the need to seek justice after harm has already occurred.