Independent Contributor, AI Governance and Human-Centred Policy

A successful first Global Dialogue on AI Governance should move beyond broad statements of principle and produce practical direction for how AI can be governed in ways that protect people, support innovation, and strengthen public trust. The Dialogue would be successful if it achieves five outcomes. First, it should establish clearer global priorities on AI safety, cybersecurity, human rights, accountability, and inclusion, especially where generative AI creates risks such as deepfakes, impersonation, misinformation, privacy violations, and digital fraud. Second, it should create space for voices beyond major technology companies and powerful states. Civil society, researchers, young professionals, developing countries, affected communities, and public-interest actors should shape the agenda, not merely respond to it. Third, it should encourage governance models that connect legal compliance, technical standards, and human-centred design. AI governance should not be treated only as a technical or regulatory exercise. It must also consider how systems affect real people in real contexts. Fourth, it should identify practical pathways for international cooperation, including shared standards for AI-enabled cyber harms, deepfake detection, cross-border enforcement, accountability reporting, and capacity-building for countries with fewer resources. Finally, it should produce a clear post-dialogue roadmap. The value of the Dialogue will depend on what happens after Geneva: what gets implemented, who is responsible, how progress is monitored, and how affected communities remain involved. In my view, success would mean leaving the Dialogue with a stronger global commitment to AI governance that is rights-based, accountable, inclusive, and usable in practice.

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

These priorities reflect the urgent need to govern AI as both a technical and human issue. Generative AI now creates risks around deepfakes, impersonation, misinformation, privacy breaches, and cyber-enabled harm. These risks affect not only digital systems, but also human dignity, autonomy, reputation, trust, and security. I selected human rights because AI governance must protect affected persons, especially where personal data, identity, voice, image, and likeness can be misused. I selected transparency, accountability, and human oversight because organisations deploying AI systems should be responsible for how those systems are used, monitored, and corrected when harm occurs. I also selected the social, economic, ethical, cultural, linguistic, and technical implications of AI because AI harms and benefits are not evenly distributed. Governance must consider real-world impact, including exclusion, inequality, language gaps, and the needs of vulnerable communities.

Yes. One emerging issue is the need to treat AI cybersecurity as a human rights and public trust issue, not only a technical security issue. Current AI governance discussions often separate safety, privacy, cybersecurity, and human rights. In practice, these issues overlap. For example, deepfakes, AI-enabled impersonation, synthetic identity fraud, and privacy violations do not only create technical risk. They can damage reputation, autonomy, dignity, democratic trust, and personal security. Another cross-cutting issue is design accountability. AI systems may satisfy formal legal or technical requirements but still harm people because they are poorly designed, difficult to challenge, or unsafe in real use. Governance should therefore examine how AI systems are experienced by affected persons, not only how they perform in controlled settings. A third issue is remedy and redress. Many AI governance conversations focus on prevention, but affected persons also need clear routes to report harm, remove harmful content, challenge misuse of their data or likeness, and obtain meaningful remedies. These issues require governance approaches that connect rights, accountability, cybersecurity, and human-centred design.

Governance gaps in AI are affecting both the UK and wider global contexts by creating a mismatch between the speed of AI adoption and the readiness of legal, institutional, and technical safeguards. The most significant challenge is that generative AI is increasing risks around deepfakes, impersonation, misinformation, privacy breaches, and cyber-enabled fraud. These harms affect individuals, public institutions, businesses, and vulnerable communities. Existing governance approaches often focus on compliance or technical risk, but do not always address how affected persons can understand, challenge, or seek remedies for AI-related harms. In my region and sector, the challenge is also one of accountability. It is often unclear who should be responsible when AI systems are misused: the developer, deployer, platform, organisation, or end user. This creates gaps in enforcement and weakens public trust. There is also an inclusion challenge. Communities with lower digital literacy, fewer resources, or limited access to legal remedies may be more exposed to AI-enabled harms and less able to respond. The opportunity is to develop governance approaches that are rights-based, practical, and human-centred. AI can improve public services, health systems, education, research, and institutional efficiency. However, this requires clear accountability, stronger cybersecurity safeguards, accessible redress systems, and design standards that protect users from the start. The key opportunity is to move AI governance from abstract principles to usable frameworks that protect people in real contexts.

The AI Dialogue can help turn fragmented AI governance efforts into a more coherent global conversation. Many countries, regions, and institutions are developing AI rules, standards, and principles, but these approaches do not always speak to one another. The Dialogue can create a shared space where governments, civil society, researchers, technical experts, and affected communities compare approaches, identify gaps, and agree on practical areas for cooperation. Its most important role should be to connect AI safety, cybersecurity, human rights, accountability, and inclusion. These issues are often discussed separately, but in practice they overlap. Deepfakes, synthetic identity fraud, privacy violations, and misinformation show why AI governance must be both technically sound and human-centred. The Dialogue can also help prevent AI governance from being shaped only by powerful states and large technology companies. It should give stronger visibility to developing countries, smaller institutions, civil society, and communities directly affected by AI harms. Its added value would be to move international cooperation from broad principles to implementation: shared standards, capacity-building, accountability mechanisms, and practical guidance for safe, rights-based AI governance.

Different stakeholders should contribute through clearly defined roles. Governments can share regulatory priorities and implementation challenges. Civil society can bring evidence of real-world harms and community impact. Technical experts can explain system risks, safety standards, and detection tools. Researchers can provide independent evidence. Private-sector actors can explain deployment practices, but should not dominate the agenda. Affected communities should help test whether proposed governance ideas work in practice. The AI Dialogue should be structured around problem-solving, not only speeches. I recommend: Thematic roundtables on safety, human rights, accountability, inclusion, and AI-enabled cyber harms. Case-based sessions using real examples such as deepfakes, fraud, bias, and public-sector AI failures. Stakeholder panels that include governments, civil society, researchers, youth, and affected communities. Practical working groups to produce short action notes after the Dialogue. Post-dialogue reporting showing what recommendations were made, who is responsible, and how progress will be tracked. The Dialogue should be inclusive, evidence-led, and implementation-focused.

Several voices remain underrepresented in global AI governance discussions. These include affected individuals, civil society groups, young researchers, Global South communities, smaller states, people with disabilities, women and girls, linguistic minorities, public-sector users, and communities with limited digital access. Too often, AI governance is shaped by governments, large technology firms, and technical experts, while those most exposed to harm have limited influence. They can be included through funded participation, regional consultations, multilingual submissions, youth and civil society panels, community evidence sessions, and accessible online participation. The Dialogue should also create space for people who have experienced AI-related harms, including privacy violations, deepfake abuse, exclusion from automated systems, or identity misuse. Inclusion should not be symbolic. These communities should help shape the agenda, test proposed solutions, and review outcomes after the Dialogue.

Design Thinking could be used as an innovative engagement format for the AI Dialogue because it would move discussions beyond formal statements into practical problem-solving. AI governance is not only a technical or legal issue; it is also a human experience issue. The Dialogue should therefore create spaces where stakeholders can test ideas around real user needs, harms, and institutional constraints. I recommend three engagement formats: 1.Design Thinking Policy Labs Mixed stakeholder groups could work through real AI governance scenarios, such as deepfake abuse, AI-enabled fraud, data misuse, or automated decision-making. Each group would identify affected users, map harms, test governance responses, and produce practical policy ideas. 2. Affected-Person Journey Mapping Participants could map what happens when a person is harmed by AI: how they detect the harm, report it, seek takedown, access remedy, and hold actors accountable. This would expose gaps that formal policy discussions often miss. 3. Prototype Governance Clinics Stakeholders could develop and test simple governance tools, such as accountability checklists, redress pathways, risk assessment templates, or rights-based design standards. This format would make the AI Dialogue more practical, inclusive, and grounded in real-world impact. It would also help convert high-level principles into usable governance mechanisms.

The OECD AI Principles provide an important global reference point for trustworthy AI that respects human rights, democratic values, transparency, robustness, and accountability. They are useful because they create a common language across countries and institutions. The UNESCO Recommendation on the Ethics of Artificial Intelligence is also important because it places human rights, dignity, fairness, transparency, and human oversight at the centre of AI governance. It is especially useful for connecting AI policy with social, cultural, and ethical impact. The Council of Europe Framework Convention on AI offers another strong model because it links AI governance directly to human rights, democracy, and the rule of law. Its rights-based approach is relevant for addressing AI-enabled harms such as privacy violations, discrimination, and misuse of personal data. The NIST AI Risk Management Framework provides a practical approach for identifying, assessing, and managing AI risks across the system lifecycle. It is useful because it helps organisations move from broad principles to operational risk management. In addition, design-led governance approaches, such as impact assessments, user journey mapping, redress pathway design, and participatory policy labs, can make AI governance more practical. These approaches help policymakers understand how AI systems affect people in real life, not only how they perform technically. Together, these examples show that effective AI governance should combine rights protection, risk management, institutional accountability, and human-centred design.