As artificial intelligence systems become increasingly advanced and autonomous, establishing minimum safety standards, independent testing, and transparent information sharing has become essential for building trust in the technology, according to United Nations Under-Secretary-General and Special Envoy for Digital and Emerging Technologies Amandeep Singh Gill.

Gill emphasized these points during his keynote address at the opening day of the 11th Singapore International Cyber Week (SICW) on October 13. The three-day event, which gathers over 15,000 participants from 90 countries and regions, focuses on the intersection of AI, cybersecurity, diplomacy, and international cooperation.

Highlighting trust as a fundamental component akin to “hard infrastructure,” Gill urged nations to establish robust AI safety frameworks. He pointed to Singapore’s AI Safety Institute, founded in 2024, as a leading example. The institute develops evaluation methodologies to detect risks associated with frontier and generative AI models, including data leaks, cybersecurity threats, fraud, and harmful content. Importantly, it assesses whether safety mechanisms function effectively across multiple languages beyond English.

Other countries, including the United Kingdom, have developed similar approaches in collaboration with private sector partners to address security concerns and provide policymakers with necessary insights. Gill stressed that these efforts must be expanded worldwide, regardless of whether a country builds, licenses, or imports AI technologies.

He also advocated for internationally recognized minimum standards for AI testing, comparable to accounting or financial benchmarks, and for certifying independent auditors to assess AI systems’ baseline behavior. “Perhaps, most importantly, we need sharing of information, incident reporting and transparency... this cannot be done through ad hoc arrangements,” Gill said, referencing a multilateral AI risk management framework proposed by UN Secretary-General Antonio Guterres in September. The framework seeks to foster global consensus on AI governance and establish mechanisms for cooperation.

Supporting this global push, Singapore joined over 20 countries and the European Union in signing a declaration calling for binding international oversight of frontier AI models to replace corporate self-regulation. During the United Nations General Assembly in New York the same month, Singapore’s Foreign Minister Vivian Balakrishnan proposed a UN framework convention specifically addressing AI safety.

Trust, according to Gill, remains one of the most significant barriers to AI adoption. He emphasized that people must understand AI systems' design and capabilities, be able to verify their claims independently, and retain human oversight to intervene when failures occur. This is especially critical for governments deploying AI in essential sectors, including those reliant on foreign-developed technologies.

Echoing the importance of trust and resilience, White House National Cyber Director Sean Cairncross highlighted the necessity for nations to fortify their cybersecurity defenses. He stressed that sovereignty in the digital age depends on protecting networks, critical infrastructure, and maintaining secure supply chains free from adversarial influence. Cairncross cautioned against relying solely on international institutions or slow global governance frameworks, advocating instead for proactive collaboration among trusted partners.

Panel discussions moderated by Singapore’s chief AI officer He Ruimin underscored the consensus among foreign officials and industry leaders on the need for robust evaluation frameworks, external testing, and clear accountability mechanisms in AI deployment. Gracie Lin, regional chief executive of crypto exchange OKX Singapore and Australia, argued that companies should leverage existing evaluations performed by developers and independent parties rather than duplicating efforts, adding that access to AI models for testing should be tailored to specific use cases.

On the question of liability when AI systems are exploited in cyberattacks, Anthropic’s head of Asia-Pacific policy, Michael Sellitto, noted that responsibility depends on context, including whether the technology was used as intended. He acknowledged the complexity of establishing a universal liability framework given AI’s wide-ranging applications, emphasizing the need for case-by-case consideration.

The discussions at SICW reflect ongoing efforts to confront the challenges posed by AI’s rapid growth and to promote responsible governance through international cooperation, transparent practices, and rigorous safety standards.