Minimum safety standards, independent testing are a must to build trust in AI: UN special envoy

As artificial intelligence models gain more autonomy, United Nations Under-Secretary-General and Special Envoy for Digital and Emerging Technologies Amandeep Singh Gill emphasized the necessity of minimum safety standards, independent testing, and open sharing of cybersecurity incidents to foster trust in AI. Speaking at the opening of the 11th Singapore International Cyber Week, Gill described trust in this AI realm as "hard infrastructure," contrasting it with the often "soft" concept of trust in general.

He urged every nation to establish a minimum safety infrastructure, highlighting Singapore's AI Safety Institute as a leading example. This institute develops evaluation methods to assess frontier and generative AI models for risks like data leaks, cybersecurity threats, fraud, and harmful content, with a focus on testing AI safety measures across various languages.

Other countries, such as the UK, have also developed AI safety infrastructure working with the private sector and policymakers. Gill called for a global expansion of these efforts, irrespective of whether a country develops, borrows, or purchases AI models. He advocated for minimum testing standards and independent auditors to verify acceptable AI baseline behavior and actions.

Transparency, including information sharing and incident reporting, is essential, and Gill pointed to UN Secretary-General Antonio Guterres' proposed multilateral AI risk management framework as an international mechanism to facilitate this. Singapore, along with over 20 countries and the European Union, signed a declaration in September advocating for binding international oversight of frontier AI models, moving away from corporate self-regulation.

Gill stressed that trust should no longer be viewed as a "soft concept," emphasizing the need for understanding, testing, and human intervention in AI systems. Public trust in AI and the ability for governments to trust systems in critical sectors are significant barriers to adoption. US White House national cyber director Sean Cairncross echoed Gill's sentiments, emphasizing trust as the foundation for America's cooperation between government and private sector in designing, building, and operating AI systems.

He called for nations to strengthen their own defenses to reinforce the broader network of allied security, emphasizing that true sovereignty in the digital age lies in the capacity to defend against adversarial influence and hidden dependencies.