Kaspersky

The Global Dialogue on AI Governance presents an important opportunity to move beyond general principles toward practical and implementable outcomes. From a cybersecurity perspective, Kaspersky believes that the following solution areas merit particular attention: • AI security: Development of comprehensible and scalable baseline practices – specifically designed for the organizations that deploy third-party AI models – for risk assessment, secure development, and adversarial testing, among others. • Standardized incident reporting: Creating a common taxonomy for AI security incidents to enable effective cross-border threat intelligence sharing. • Interoperability: Greater alignment and mapping across national regulatory approaches and technical standards to reduce fragmentation. • Capacity building: Targeted support for developing countries in building secure AI infrastructure, skills, and governance capabilities through practical guidelines, courses, and workshops. • Transparency and accountability mechanisms: Practical tools such as audit frameworks, documentation standards, and third-party assurance models. • Vulnerability coordination: Exploration of a mechanism for responsible disclosure of vulnerabilities in widely deployed AI models. To ensure impact, the Co-Chairs' Summary should include areas of convergence and divergence between stakeholders, applied priority actions, indicative pathways for cooperation and follow-up, as well as indicators to assess progress over time.

Safe, secure and trustworthy AI;AI capacity-building;Interoperability of governance approaches;Transparency, accountability, and human oversight;

As a global cybersecurity company, Kaspersky believes that the Global Dialogue on AI Governance should aim to foster responsible AI development, global cooperation, and ethical standards, ensuring outcomes that are actionable and aligned with global objectives of safety, inclusivity, and sustainability in AI technologies. Specifically, the Dialogue could contribute to establishing internationally recognized standards and guidelines for AI development and deployment, focusing on ethics, security, transparency, and accountability. It should also work towards creating cybersecurity protocols and best practices for AI systems, aimed at mitigating risks such as data breaches, malicious attacks, and vulnerabilities in critical infrastructure. The selected thematic areas are intricately intertwined, and their prioritization could help achieve the abovementioned objectives, encouraging a substantive multistakeholder dialogue.

In our view, the proposed thematic clusters provide a coherent high-level framework aligned with current international discussions on AI governance. They appropriately capture key dimensions, including security considerations, capacity gaps, and ethical impacts. However, from an operational and implementation perspective, further refinement would enhance clarity and accessibility across stakeholder groups. Specifically, some overlap exists between the clusters on "safe, secure and trustworthy AI" and "human rights, transparency and accountability," particularly regarding oversight, assurance, and audit mechanisms. In this regard, greater delineation would improve usability. In addition, from our perspective, cybersecurity risks - such as adversarial attacks, model integrity, supply chain vulnerabilities, and data poisoning - should be more explicitly reflected, especially within the "safe, secure and trustworthy AI" cluster.

To ensure inclusive, effective, and representative decision-making, all relevant stakeholders must be actively engaged in the Global Dialogue on AI Governance. This should be an ongoing process involving continuous discussions, regular meetings, and updates. A broad multistakeholder community - including governments, the private sector, civil society, academic institutions, and international organizations - is essential for ensuring that AI governance is both holistic and comprehensive.

Governments play a key role in shaping AI policies, drafting governance frameworks, and ensuring alignment between national regulations and global standards. Additionally, governments should work together and with other stakeholders to prevent fragmented policies across borders and safeguard fundamental human rights. The private sector, particularly AI developers, tech companies, and cybersecurity firms, must also be involved to ensure the dialogue reflects the realities of industry and the pace of technological change. Tech companies can contribute by offering insights into the development and deployment of AI, sharing expertise on AI-driven risks, providing best practices for secure AI use, and offering perspectives on how AI policies impact innovation and business operations. Such knowledge sharing could be facilitated through reports and whitepapers, thematic workshops, roundtables, and plenary sessions featuring other relevant stakeholders. Kaspersky advocates for a more active involvement of governments in collaboration with non-state actors within the Dialogue. To facilitate deeper exploration of specific AI governance topics, thematic expert panels and workshops should be organized, drawing on the expertise of specialists and practitioners. This would promote active participation from industry stakeholders and help generate practical, actionable outcomes. It is also essential to maintain broad participation through open consultations, ensuring that diverse voices are included in the process.

To maximize substantive engagement, particularly on security matters, we propose consider replacing one high-level governmental plenary session with – or reserving a specific timeslot for – a moderated, interactive roundtable featuring national cybersecurity entities (CERTs/CSIRTs) and leading industry stakeholders to share insights on current AI threat landscapes. Such an exchange would ground the high-level dialogue in operational, solution-oriented reality.