Forti5 Tech Ltd, UK

Established governance globally

• Safe, secure and trustworthy AI
• AI capacity-building
• Social, economic, ethical, cultural, linguistic and technical implications of AI
• Interoperability of governance approaches

Building safe secure and trustworthy must be done as by design and AI companies should show a complete transparency of their established by design life cycle process.

AI by Design principles

By Design process and standardization

Establishing AI security, ethics, and transparency by design standardised life cycle process.

A uniform approach

In a structured manner, reference Ramachandran, M (2025) Engineering AI Ethics by Design, Springer, https://link.springer.com/book/10.1007/978-981-95-2909-4

Meaningful AI Dialogue requires structured, inclusive participation across governments, civil society, academia, industry, and affected communities. We recommend the following framework to make such participation equitable and effective. Stakeholder Roles Governments should provide policy context and regulatory intent, while remaining open to technical challenge from industry and academia. Civil society and community representatives must have dedicated, compensated roles — not token participation — to ensure the Dialogue reflects those most affected by AI deployment. Academic institutions should contribute independent evidence and cross-disciplinary analysis. Industry should share real-world implementation experience, including honest disclosure of failure modes and unresolved safety challenges. Recommended Format and Structure ∙ Tiered participation model: open written submissions for all stakeholders, with selected participants invited to structured thematic working sessions. ∙ Dedicated tracks for underrepresented voices: including Global South nations, Indigenous communities, and small and medium enterprises, with translation and access support provided. ∙ Pre-Dialogue synthesis process: an independent rapporteur function should consolidate written submissions into evidence briefs before plenary sessions, so Dialogue time is spent on deliberation, not repetition. ∙ Iterative feedback loops: draft conclusions should be published in interim form with a structured stakeholder review period before finalisation. ∙ Transparent contribution registry: all substantive inputs should be publicly attributed and traceable to outcomes, supporting accountability and discouraging capture by well-resourced actors. AI governance decisions made without broad stakeholder input risk embedding the values and risk tolerances of a narrow set of actors. Structure shapes legitimacy: the format of the Dialogue is itself a governance choice.

Stakeholder Roles Static panel presentations and formal plenary sessions have limited value in a fast-moving domain like AI. The UN AI Dialogue should adopt engagement formats that are participatory, evidence-driven, and designed to surface genuine disagreement as well as consensus. Recommended Innovative Formats ∙ Live scenario stress-testing: Present real-world AI deployment scenarios — such as autonomous decision-making in public services, or AI-assisted hiring — and invite multi-stakeholder groups to identify risks, gaps, and governance responses in real time. This surfaces practical tensions that abstract policy discussion often misses. ∙ Red team / blue team sessions: Structured adversarial formats where one group argues for a proposed AI governance measure and another challenges it. This prevents echo chambers and stress-tests policy proposals before they are adopted. ∙ Citizen deliberative panels: Randomly selected members of the public, supported by expert briefings, deliberate on specific AI governance questions and present conclusions to the Dialogue. This has proven effective in climate and constitutional processes and brings legitimacy beyond organised stakeholder groups. ∙ Open evidence walls: A continuously updated, publicly visible repository of submitted evidence, case studies, and technical assessments that delegates can reference and challenge throughout the Dialogue — not just in formal sessions. ∙ Cross-regional peer exchanges: Structured bilateral or small-group sessions pairing delegations from different regions to compare implementation experiences, reducing the dominance of a single regulatory model. ∙ Async digital participation layer: A structured online platform running in parallel with in-person sessions, allowing stakeholders unable to attend physically to contribute questions, vote on priorities, and respond to draft conclusions in real time. Formats drive outcomes. A Dialogue designed around listening will produce richer, more legitimate governance frameworks than one designed around presentations.

Effective AI governance requires more than principles - it requires implementable frameworks, tested practices, and platforms that translate policy intent into operational reality. The following examples represent approaches that have demonstrated measurable value. Policy Frameworks ∙ EU AI Act (2024): The first binding risk-tiered regulatory framework for AI, categorising systems by risk level and imposing proportionate obligations. Its conformity assessment and transparency requirements offer a replicable model for other jurisdictions. ∙ NIST AI Risk Management Framework (AI RMF 1.0): A voluntary but widely adopted framework providing structured guidance on governing, mapping, measuring, and managing AI risk across organisations. Its sector-agnostic design makes it adaptable internationally. ∙ UNESCO Recommendation on the Ethics of AI (2021): Adopted by 193 member states, this provides a shared normative foundation including principles on human oversight, transparency, and environmental sustainability of AI systems. Practices ∙ Algorithmic impact assessments: Modelled on environmental impact assessments, these require organisations to evaluate potential harms before deploying AI systems in high-stakes contexts such as criminal justice, welfare, and healthcare. ∙ Responsible AI procurement standards: Several governments, including the UK and Canada, have introduced supplier requirements mandating transparency, explainability, and human oversight for AI systems procured for public use. ∙ AI incident registries: Public databases of AI failures and near-misses, analogous to aviation safety reporting, enable cross-sector learning and evidence-based policy refinement. Platforms ∙ OECD AI Policy Observatory: Provides comparative, evidence-based analysis of national AI strategies and regulatory approaches, supporting policy coherence across jurisdictions. ∙ Partnership on AI: A multi-stakeholder platform bringing together civil society, academia, and industry to develop shared best practices and publish research on responsible AI deployment. Governance effectiveness depends on implementation infrastructure, not only on normative agreement.