Last week, the UN General Assembly adopted by consensus a US-led resolution on “Seizing the opportunities of safe, secure and trustworthy artificial intelligence systems for sustainable development,” backed by more than 120 other Member States. Earlier this year, TFS submitted written comments in response to previous drafts of the resolution, several of which are reflected in the adopted resolution.

In early January, TFS was invited to provide feedback on a draft US-led United Nations (UN) General Assembly resolution on “Seizing the opportunities of safe, secure and trustworthy artificial intelligence systems for sustainable development.” Our recommendations strived to ensure transparency and accountability in the development of advanced AI systems, enhance cybersecurity requirements for developers of advanced AI systems, and improve regulatory harmony.

TFS was then invited by the U.S. State Department to present our recommendations at a roundtable attended by a select group of participants representing international civil society organizations, academia, and other non-governmental groups.

Last week, the UN General Assembly adopted by consensus the US-led resolution, backed by more than 120 other Member States. We are elated to see that the finalized resolution contains amendments from earlier versions, reflecting some of our recommendations (emphasized in bold):

• Risk prevention: “Encouraging effective measures, that promote innovation for the internationally interoperable identification, classification, evaluation, testing, prevention and mitigation of vulnerabilities and risks during the design and development and prior to the deployment and use of artificial intelligence systems…” (Section 6b)
• Reporting incidents and misuse: “Encouraging the incorporation of feedback mechanisms to allow evidence-based discovery and reporting by end-users and third parties of technical vulnerabilities and, as appropriate, misuses of artificial intelligence systems and artificial intelligence incidents following their development, testing, and deployment to address them…” (Section 6c)
• Risk monitoring and impact assessments: “Fostering the development, implementation and disclosure of mechanisms of risk monitoring and management, mechanisms for securing data, including personal data protection and privacy policies, as well as impact assessments as appropriate, across the life cycle of artificial intelligence systems” (Section 6e)
• Bias in datasets: “Facilitating the development and implementation of effective, internationally interoperable frameworks, practices and standards for training and testing artificial intelligence systems to enhance policymaking and to help protect individuals from all forms of discrimination, bias, misuse or other harm, and avoid reinforcing or perpetuating discriminatory or biased applications and outcomes throughout the life cycle of artificial intelligence systems, including, for example, by analysing and mitigating bias encoded in datasets and otherwise combating algorithmic discrimination and bias, while not inadvertently or disproportionally impacting the positive development, access and uses of other users and beneficiaries…” (Section 6h)
• Open-source AI systems: “Encouraging research and international cooperation to understand, balance and address the potential benefits and risks related to the role of artificial intelligence systems in bridging digital divides and achieving all 17 Sustainable Development Goals, including the role of scaling up of digital solutions such as open-source artificial intelligence systems…” (Section 6o)
• Inclusivity in AI research and development: “Acknowledges the importance of … continued efforts by the international community to promote inclusive research…”(Section 8)

More details about the resolution and vote can be found in the UN News announcement.

Photography by Loey Felipe.