Collective Impact Coalition for Ethical AI (by the World Benchmarking Alliance)

We would like to see the Global Dialogue tackle the following topics within its proposed structure. 1. Promoting effective AI impact assessments by companies (thematic cluster: Safe, secure and trustworthy AI: responsible and interoperable approaches) As responsible investors investor participants in WBA's Collective Impact Coalition (CIC) for Ethical AI, we strongly believe the AI governance dialogue should include discussions on AI systems impact assessments by technology companies, to ensure risks and harms are avoided. 2. Strengthen clarity around human oversight of AI (thematic cluster: Safe, secure and trustworthy AI: responsible and interoperable approaches) Human oversight of AI governance implementation is embedded in key expectations across most responsible AI frameworks. However, as investors we observe that translating policy into practice – and reporting on this aspect of responsible AI implementation – remains a challenge for companies and hence a barrier to investors' and other stakeholders' ability to assess the extent of companies' AI risk management. The Global Dialogue should include discussions of how companies can implement effective oversight of AI and communicate on this transparently. 3. Meaningful and effective policy enforcement and harmonization of rules (thematic cluster: Respecting, protecting and promoting human rights: transparency, accountability and human oversight) Without enforceable transparency and accountability standards, AI risks will continue to grow unchecked. Therefore, we would advocate for the topic of policy enforcement to be included I the dialogue program. Policymakers must act decisively to close the accountability gap in AI governance by mandating robust, transparent HRIAs across the technology sector. First, legislation should require all companies developing or deploying AI systems to conduct comprehensive, standardised HRIAs that clearly define scope, affected populations, intended uses, and potential risks. These assessments must not remain internal documents; public disclosure should be compulsory, with accessible summaries that enable scrutiny by regulators, civil society, and affected communities.

• Safe, secure and trustworthy AI
• Protection and promotion of human rights
• Transparency, accountability, and human oversight
• AI capacity-building

We would like to see the Global Dialogue tackle the following topics within its proposed structure. 1. Promoting effective AI impact assessments by companies (thematic cluster: Safe, secure and trustworthy AI: responsible and interoperable approaches) As responsible investors investor participants in WBA's Collective Impact Coalition (CIC) for Ethical AI, we strongly believe the AI governance dialogue should include discussions on AI systems impact assessments by technology companies, to ensure risks and harms are avoided. 2. Strengthen clarity around human oversight of AI (thematic cluster: Safe, secure and trustworthy AI: responsible and interoperable approaches) Human oversight of AI governance implementation is embedded in key expectations across most responsible AI frameworks. However, as investors we observe that translating policy into practice - and reporting on this aspect of responsible AI implementation - remains a challenge for companies and hence a barrier to investors' and other stakeholders' ability to assess the extent of companies' AI risk management. The Global Dialogue should include discussions of how companies can implement effective oversight of AI and communicate on this transparently. 3. Meaningful and effective policy enforcement and harmonization of rules (thematic cluster: Respecting, protecting and promoting human rights: transparency, accountability and human oversight) Without enforceable transparency and accountability standards, AI risks will continue to grow unchecked. Therefore, we would advocate for the topic of policy enforcement to be included I the dialogue program. Policymakers must act decisively to close the accountability gap in AI governance by mandating robust, transparent HRIAs across the technology sector. First, legislation should require all companies developing or deploying AI systems to conduct comprehensive, standardised HRIAs that clearly define scope, affected populations, intended uses, and potential risks. These assessments must not remain internal documents; public disclosure should be compulsory, with accessible summaries that enable scrutiny by regulators, civil society, and affected communities.

No

As the new generation of AI tools continue to advance, they could severely distort public discourse, amplify misinformation, facilitate arbitrary surveillance and act as vehicles for discrimination and gender-based violence. Transparent assessments on the impacts of AI are necessary to avoid harms. Without thorough assessments of AI systems and tools, these risks will fall through the cracks and escalate into actual harms. According to the latest findings by the World Benchmarking Alliance, none of the 200 leading digital technology companies showed proof of conducting comprehensive human rights impact assessments (HRIAs) on the AI systems they developed, procured or deployed, which is alarming. Even companies that claimed to conduct appropriate HRIAs rarely disclosed their scope, affected groups, AI functionalities covered or assessment results. All of these are essential to robust accountability and are enshrined in both guidance from human rights institutes and emerging legislation. As it stands, only Microsoft's commissioned assessment of technologies licensed to law enforcement agencies comes close to meeting WBA's strengthened assessment criteria.

Harmonisation of expectations across jurisdictions is important for enforcement and to avoid global policy fragmentation which could create ambiguity for companies and investors alike. The Global Dialogue provides an excellent platform to facilitate dialogue between international policymakers and stakeholders to support this process.

The UN Global Digital Compact

The UN should consider organising regional dialogues/consultation sessions to ensure inclusivity in the process. These could be through virtual and physical workshops to collect input from stakeholders.

People who have no acces sto the internet or innovations. The dialogue should ensure includivity of their voices.

Policymakers must act decisively to close the accountability gap in AI governance by mandating robust, transparent HRIAs across the technology sector. First, legislation should require all companies developing or deploying AI systems to conduct comprehensive, standardised HRIAs that clearly define scope, affected populations, intended uses, and potential risks. These assessments must not remain internal documents; public disclosure should be compulsory, with accessible summaries that enable scrutiny by regulators, civil society, and affected communities. Second, governments should establish independent oversight bodies with the authority to audit AI systems and verify the quality and completeness of reported assessments. Voluntary compliance has proven insufficient. Binding enforcement mechanisms, including penalties for non-disclosure or inadequate assessments, are essential. Third, policymakers should ensure that HRIAs are continuous, not one-off exercises. AI systems evolve rapidly, and so should the evaluation of their societal impacts. Requirements for periodic reassessment and real-time risk monitoring can help prevent harm before it escalates. As well, meaningful stakeholder engagement must be embedded in the process. Communities most at risk of discrimination, surveillance, or misinformation should have a voice in shaping assessments and mitigation strategies. Policymakers should consider how these expectations can be implemented across different market contexts, particularly in emerging markets where regulatory or technical capacity may still be developing. In such contexts, a phased or context-sensitive approach may be necessary to ensure that requirements remain achievable while still upholding core accountability principles. This could include tiered obligations based on risk level, targeted technical assistance, and the development of shared tools or templates to support implementation.