Hacettepe University / UNESCO Women for Ethical AI

The first Global Dialogue on AI Governance will be a success if it produces one thing that previous AI governance processes have consistently failed to deliver: a clear, binding framework that positions corporations as duty-bearers — not merely as stakeholders. The AI systems reshaping labor markets, healthcare, public services, and democratic processes are not designed by states. They are designed, deployed, and monetized by corporations. Yet the dominant architecture of international AI governance continues to treat corporate actors as participants in dialogue rather than subjects of obligation. This is not a gap — it is a structural choice, and it has consequences. A successful first Dialogue would establish three concrete outcomes: First, agreement that human rights due diligence for AI — grounded in the UN Guiding Principles on Business and Human Rights — must be extended, strengthened, and made operationally specific for AI systems across the full development and deployment lifecycle. Second, recognition that compliance-based corporate accountability is insufficient. A corporation can satisfy every regulatory checkpoint and still deploy systems that systematically erode human dignity, autonomy, and opportunity. The Dialogue must begin building consensus around a human value standard — one that asks not only whether rules were followed, but whether human capabilities were protected. Third, the establishment of a dedicated working stream on corporate AI accountability within the Dialogue's ongoing architecture, ensuring that private sector actors engage as duty-bearers in 2027 and beyond — not only as technical advisors or industry representatives. Governance without accountability is architecture without walls. The first Dialogue's lasting contribution will be measured by whether it begins to close the structural gap between the power corporations exercise over AI and the obligations they currently bear.

• Transparency, accountability, and human oversight
• Social, economic, ethical, cultural, linguistic and technical implications of AI
• Interoperability of governance approaches
• Protection and promotion of human rights

My four priorities - protection and promotion of human rights, transparency and accountability, social and ethical implications of AI, and interoperability of governance approaches - form a single coherent argument: effective AI governance requires corporate actors to be structurally accountable, not merely engaged. Human rights is the foundational priority. AI systems deployed at scale are not neutral tools - they make consequential decisions about people's access to employment, credit, healthcare, education, and justice. These decisions carry human rights implications that existing frameworks address inadequately. The UN Guiding Principles on Business and Human Rights provide the correct foundation, but require AI-specific elaboration and, critically, stronger operationalization at the corporate level. Transparency and accountability cannot be reduced to technical explainability. Accountability means that when an AI system causes harm, there is an identifiable responsible party, a clear obligation to remedy, and a mechanism to enforce it. Currently, accountability dissolves across the AI value chain - between developers, deployers, and users. The Dialogue must address this structural diffusion directly. Social and ethical implications extend beyond bias and fairness. They include the structural capacity of AI systems to reshape labor markets, redistribute economic power, and alter the conditions under which people exercise fundamental freedoms. Addressing these implications requires engagement with corporations as the primary agents of these transformations. Interoperability of governance approaches matters precisely because corporations operate across jurisdictions. Without minimum corporate accountability standards embedded within interoperability frameworks, regulatory divergence becomes an incentive for regulatory arbitrage - companies deploying their most consequential systems in the jurisdictions with the weakest protections. These four priorities are interconnected. Addressing any one of them in isolation, without corporate accountability as the binding thread, will produce compliance documentation rather than human protection.

The listed themes, while comprehensive in scope, share a structural omission: none explicitly addresses the question of who is obligated within AI governance frameworks. This is not a thematic gap - it is a foundational one. Current international AI governance is predominantly state-addressed. Norms are directed at governments; interoperability frameworks assume state-level coordination; capacity-building programs target public institutions. Yet the entities most consequentially shaping AI's impact on human lives are private corporations - and the governance architecture consistently positions them as contributors to dialogue rather than subjects of obligation. This produces what I identify in my doctoral research as a structural absence: the most powerful actors in the AI ecosystem operate in a space where international human rights norms apply to states, domestic regulation varies by jurisdiction, and corporate self-governance fills the remaining space with voluntary commitments. The result is sophisticated governance language with limited accountability infrastructure. Two cross-cutting issues require urgent attention: Corporate AI accountability as a distinct governance subject. The Business and Human Rights framework has spent two decades attempting to bring corporations into binding accountability structures. The AI Dialogue has an opportunity - and an obligation - to build on that trajectory rather than reproduce its limitations. This means moving beyond due diligence as a reporting exercise toward enforceable human rights obligations for AI developers and deployers. The standard of evaluation. Governance frameworks consistently measure corporate behavior against compliance benchmarks: did the company follow applicable rules? This is necessary but insufficient. A human value standard - grounded in whether AI systems expand or erode the conditions for human flourishing - demands a structurally different form of corporate accountability. The Dialogue should begin building consensus around this higher standard. Without addressing who is obligated and to what standard, the Dialogue risks producing consensus on principles while leaving the accountability gap intact.

The most significant governance challenge facing emerging economies with substantial digital infrastructure is not technological — it is structural. These states are not the primary drivers of AI development. They are deployment destinations. AI systems designed, trained, and governed according to the standards of a small number of technologically advanced jurisdictions are deployed at scale into societies that had no role in shaping those systems and possess limited institutional capacity to contest their consequences. This asymmetry is compounded by a deeper power imbalance. The corporations driving AI development command data assets, computational resources, and technical expertise that exceed the regulatory capacity of most states. Governments that lack the infrastructure to audit, challenge, or meaningfully oversee these systems are not governing AI — they are accommodating it. This is not a capacity problem that training programs alone can resolve. It is a structural power imbalance that governance frameworks must explicitly name and address. Domestically, many such states lack binding AI governance frameworks, leaving corporate actors to self-regulate according to voluntary principles or comply selectively with extraterritorial instruments designed for different institutional contexts. The result is a paradox: companies face increasing regulatory complexity while citizens enjoy decreasing accountability. From a research and practitioner perspective — informed by over sixteen years working at the intersection of corporate AI governance, human rights, and technology policy, and by teaching both Business and Human Rights and AI, Ethics and Human Rights at university level — the pattern is consistent. Corporate AI governance in deployment-destination countries defaults to the lowest available standard. Without binding international obligations that apply to corporations regardless of where they are headquartered, this gap is not incidental — it is structural and self-reinforcing. The Global Dialogue must reframe AI governance as a binding architecture that follows systems — not headquarters.

The AI Dialogue arrives at a moment when the international community faces a fundamental choice: whether AI governance will be built around state cooperation on principles, or around binding accountability for the actors who actually shape AI's impact on human lives. Most existing international AI governance initiatives have chosen the first path. They produce principles, recommendations, and voluntary frameworks that states endorse and corporations selectively adopt. The result is a governance landscape that is architecturally sophisticated and accountability-thin. The AI Dialogue can and must choose differently. The unique role the Dialogue can play is to serve as the first genuinely universal forum in which corporations are engaged not as stakeholders offering technical expertise, but as duty-bearers with human rights obligations. This is not a radical departure from existing international law — it is the logical extension of it. The UN Guiding Principles on Business and Human Rights established that corporations bear responsibility to respect human rights. The UN Working Group on Business and Human Rights has progressively developed that framework. What has been missing is a mechanism that applies this architecture specifically and systematically to AI. The Dialogue is positioned to fill that gap — but only if it resists the institutional tendency to prioritize consensus on language over clarity on obligation. International cooperation on AI governance is not advanced by producing another set of principles that powerful actors can sign without consequence. It is advanced by establishing the normative foundation for corporate AI accountability that is universal in application, human rights-based in standard, and enforceable in design. The measure of this Dialogue's contribution to international cooperation will not be the breadth of participation it achieves — it will be the depth of obligation it establishes.

The AI Dialogue does not need to construct AI governance architecture from the beginning. The foundational work exists. The critical task is to connect, extend, and where necessary, strengthen what has already been built — with corporate accountability as the binding thread. The most important existing mechanism is the UN Working Group on Business and Human Rights, which has already begun engaging with technology corporations as human rights duty-bearers. The normative foundation it has developed — grounded in the UN Guiding Principles on Business and Human Rights — provides the most legally coherent and internationally recognized framework for extending corporate obligation into the AI domain. The Dialogue should formally engage the Working Group rather than treat AI governance as a separate normative universe. The UNESCO Recommendation on the Ethics of Artificial Intelligence provides the broadest intergovernmental consensus on AI ethics to date, with human rights at its center. The OECD AI Principles offer the most developed intergovernmental standard on trustworthy AI. The Council of Europe Framework Convention on AI and Human Rights (CETS No. 225) represents the first binding international instrument — limited in geographic scope but significant in normative direction. These three instruments together establish that the international community has already moved beyond principles toward standards. The Dialogue's added value is to universalize that trajectory. What none of these frameworks delivers — and what the Dialogue uniquely can — is a universally applicable corporate accountability standard that is not jurisdiction-dependent. The EU AI Act applies territorially. The UNGPs remain voluntary. CETS No. 225 binds only its signatories. The Dialogue operates within the UN system, with the broadest possible membership and the highest possible legitimacy. That combination — universal reach, human rights foundation, corporate accountability focus — is the Dialogue's irreplaceable added value.

The AI Dialogue's structure must reflect a fundamental distinction that current AI governance forums consistently blur: the difference between stakeholders who participate in governance and actors who bear obligations within it. Civil society organizations, academic researchers, technical experts, and affected communities all have legitimate and necessary roles as participants — providing evidence, representing interests, and holding the process accountable. But corporations occupy a categorically different position. They are not merely participants in AI governance. They are the primary agents of AI deployment and, consequently, primary duty-bearers under international human rights law. Treating them as one stakeholder category among many structurally obscures this distinction. The Dialogue's format should therefore establish two distinct corporate engagement tracks. The first is a technical and advisory track, where industry actors contribute expertise on system design, risk assessment, and implementation challenges — a legitimate and necessary contribution. The second is an accountability track, where corporations with significant AI deployment present their human rights due diligence records, respond to questions from affected community representatives and civil society, and engage with binding standard-setting discussions as obligation-holders rather than as advisors. This structural separation is not adversarial — it is architecturally honest. It reflects the reality that the same actor cannot simultaneously define governance standards and be the primary subject of those standards without creating an inherent conflict of interest. Additionally, the Dialogue should establish a permanent civil society and affected communities mechanism with guaranteed access, dedicated speaking time, and formal input into outcome documents — not as a symbolic inclusion gesture, but as a structural accountability feature. Format follows function. If the Dialogue's function is genuine accountability, its structure must make accountability possible.

The most consequentially underrepresented voices in global AI governance are not those of governments or industries that lack technical capacity. They are the voices of people who bear the direct, daily consequences of AI deployment — and who have had no role in designing, governing, or contesting the systems that shape their lives. Workers whose employment prospects are screened, ranked, or eliminated by automated systems. Communities subjected to algorithmic surveillance in public spaces. Individuals denied access to credit, healthcare, or social services by decision systems they cannot see, challenge, or appeal. Migrants and refugees processed through automated identification systems with no avenue for human review. These are not marginal cases — they represent the lived reality of AI governance failure at scale. Their absence from governance dialogues is not accidental. Participation in international forums requires resources, institutional affiliation, technical literacy, and access to information that affected communities systematically lack. The Dialogue must address this structurally, not rhetorically. As a member of UNESCO Women for Ethical AI, I would also highlight the persistent underrepresentation of women — particularly women from the Global South — in AI governance processes at every level: as technical developers, as policymakers, and as rights-holders whose specific experiences of AI-driven harm remain inadequately documented and addressed. From my experience teaching AI, Ethics and Human Rights and Business and Human Rights at university level, I observe a further gap: the next generation of professionals entering AI-shaped labor markets is largely absent from governance conversations about those markets. Young people, students, and early-career workers have direct stakes in AI governance outcomes and distinctive perspectives that current processes do not systematically capture. Inclusion is not achieved by adding seats at an existing table. It requires redesigning the table around the question of whose protection governance is ultimately for.

The most significant innovation the AI Dialogue could introduce is not technological — it is procedural. The format innovation that would most meaningfully advance the Dialogue's accountability function is the introduction of structured corporate human rights due diligence hearings. In this format, corporations with significant AI deployment — particularly those whose systems affect large populations across multiple jurisdictions — would present their human rights due diligence records before a panel that includes affected community representatives, independent human rights experts, and civil society organizations. The presentation would not be promotional. It would be evaluative: what harms were identified, what remediation was undertaken, what systemic changes were implemented. This format exists in analogous contexts. Parliamentary scrutiny hearings, UN Universal Periodic Review processes, and OECD National Contact Point procedures all demonstrate that structured accountability dialogues between powerful actors and rights-holders are institutionally feasible. The AI Dialogue should adapt this architecture for the AI-specific context. Beyond this central innovation, three complementary format recommendations would strengthen meaningful engagement: Affected community testimony sessions — dedicated time within the formal programme for individuals and communities directly impacted by AI systems to present evidence of harm and governance failure, with guaranteed inclusion in outcome documents. Pre-Dialogue written submission synthesis — a publicly available, independently prepared synthesis of written submissions organized by thematic area and stakeholder category, ensuring that inputs from non-institutional actors receive the same analytical weight as state and corporate contributions. Accountability gap tracking — a mechanism that carries unresolved accountability questions from the 2026 Dialogue into the 2027 session, creating continuity and preventing the pattern common to multilateral forums where difficult questions are acknowledged but not pursued. Format is not procedural detail. It determines whose voice shapes outcomes and whose experience remains invisible.

Effective AI governance already exists in partial form across multiple instruments and levels. The challenge is not the absence of good practice - it is the absence of a universal, binding architecture that connects these practices into a coherent accountability framework applicable to corporations regardless of jurisdiction. The Council of Europe Framework Convention on Artificial Intelligence and Human Rights, Democracy and the Rule of Law (CETS No. 225) represents the most significant advance in binding AI governance to date. It demonstrates that legally binding international instruments on AI are institutionally achievable. Its limitation - geographic scope restricted to Council of Europe members and invited signatories - is precisely the gap the Global Dialogue must address by extending its normative direction universally. The EU AI Act demonstrates both the potential and the structural limits of jurisdiction-dependent regulation. It establishes risk-based corporate obligations and enforcement mechanisms that represent genuine governance progress. It also creates regulatory arbitrage incentives: corporations may deploy their most consequential systems in jurisdictions with weaker frameworks. Effective global governance must close this arbitrage space. The UN Guiding Principles on Business and Human Rights, and specifically their Pillar II operationalization through human rights due diligence practices, provide the most legally coherent existing framework for corporate AI accountability. Leading corporations have begun integrating AI-specific human rights impact assessments into their governance processes. This practice must move from voluntary adoption to universal obligation. The UNESCO Recommendation on the Ethics of Artificial Intelligence, supported by its Readiness Assessment Methodology, offers the broadest intergovernmental consensus instrument currently in use - providing governments and corporations with a practical implementation framework grounded in human rights principles. Together these instruments point in the same direction. The Global Dialogue's contribution is to complete the architecture they have begun.