As artificial intelligence continues to advance rapidly in capability and deployment, new risks and potential misuse are also emerging, particularly with increasingly powerful frontier models. There is a need to accelerate investments in AI safety and alignment research and development (R&D) to keep pace with commercially driven advancements in AI capabilities. Global cooperation among countries will be necessary to optimise limited resources, maximise collective impact, and ensure that AI systems remain aligned with human intentions and societal values.
In recognition of this, the inaugural International Scientific Exchange (ISE) on AI Safety was convened in May 2025, bringing together leading scientists and practitioners to advance global collaboration in AI safety research. A key outcome was the Singapore Consensus on Global AI Safety Research Priorities (Singapore Consensus), which presents a shared global understanding of the AI safety research problem statements considered to be of highest priority.
One year on, the AI landscape has evolved significantly. As highlighted in the 2026 International AI Safety Report (published in February 2026), there has been significant improvement in capabilities due to new training methods such as distillation. Agent deployments are anticipated to increase rapidly. At the same time, there has been growing misuse and incidents involving AI, including AI-generated sexual content, AI-enabled scams, and cyber-attacks.
ISE 2026 (held from 17 to 19 May 2026) once again convened the top global AI experts, to update the Singapore Consensus to reflect these new developments, identify new priorities, and advance global collaboration in this space.
.webp)
