Global Scientific Dialogue on Frontier AI Safety
Global Scientific Dialogue on Frontier AI Safety
Global Scientific Dialogue on Frontier AI Safety
Global Scientific Dialogue on Frontier AI Safety
PAST EVENT
February 19, 2026 | 2:30PM to 6:00PM The Imperial Hotel, New Delhi Co-organised with the Brookings Institution | Supported by Minderoo Foundation and Future of Life Institute
This closed-door roundtable was convened under Chatham House Rule. In accordance with this rule, the summary below describes the themes and findings discussed without attributing views to individual participants.
AI Safety Connect and the Brookings Institution convened approximately 30 AI safety researchers, governance actors, and technical advisors from India, Japan, Asia-Pacific, the United States, the United Kingdom, and Europe for a frank, off-the-record exchange on scientific priorities and realistic coordination pathways for frontier AI safety.
AI Safety Connect and the Brookings Institution convened approximately 30 AI safety researchers, governance actors, and technical advisors from India, Japan, Asia-Pacific, the United States, the United Kingdom, and Europe for a frank, off-the-record exchange on scientific priorities and realistic coordination pathways for frontier AI safety.
AI Safety Connect and the Brookings Institution convened approximately 30 AI safety researchers, governance actors, and technical advisors from India, Japan, Asia-Pacific, the United States, the United Kingdom, and Europe for a frank, off-the-record exchange on scientific priorities and realistic coordination pathways for frontier AI safety.
Low-Stakes Coordination (3:15–4:05PM)
Two provocations framed table discussions on what safety mechanisms should be applied within the next 12 months and what safety-relevant research institutions should share openly. The session examined the geopolitical landscape for AI safety coordination, the distinction between misuse risks and loss-of-control risks, and the potential for reframing AI safety as AI reliability to unlock commercial cooperation. A concrete proposal emerged: dedicating at least 1% of AI capability investment to building AI-driven active defense systems.
Two provocations framed table discussions on what safety mechanisms should be applied within the next 12 months and what safety-relevant research institutions should share openly. The session examined the geopolitical landscape for AI safety coordination, the distinction between misuse risks and loss-of-control risks, and the potential for reframing AI safety as AI reliability to unlock commercial cooperation. A concrete proposal emerged: dedicating at least 1% of AI capability investment to building AI-driven active defense systems.
Two provocations framed table discussions on what safety mechanisms should be applied within the next 12 months and what safety-relevant research institutions should share openly. The session examined the geopolitical landscape for AI safety coordination, the distinction between misuse risks and loss-of-control risks, and the potential for reframing AI safety as AI reliability to unlock commercial cooperation. A concrete proposal emerged: dedicating at least 1% of AI capability investment to building AI-driven active defense systems.
Realistic Research and Application Pathways
(4:20–5:20PM)
Three provocations framed discussion on missing coordination infrastructure, preparation for follow-up dialogue, and differences in national AI risk assessments. The session confronted the gap between evaluation and assurance, the rapid obsolescence of safety benchmarks, and the capacity gap facing Global South countries. Participants proposed defining a minimal assurance stack within six months and establishing a dedicated Safety Benchmark Foundation.
Three provocations framed discussion on missing coordination infrastructure, preparation for follow-up dialogue, and differences in national AI risk assessments. The session confronted the gap between evaluation and assurance, the rapid obsolescence of safety benchmarks, and the capacity gap facing Global South countries. Participants proposed defining a minimal assurance stack within six months and establishing a dedicated Safety Benchmark Foundation.
Three provocations framed discussion on missing coordination infrastructure, preparation for follow-up dialogue, and differences in national AI risk assessments. The session confronted the gap between evaluation and assurance, the rapid obsolescence of safety benchmarks, and the capacity gap facing Global South countries. Participants proposed defining a minimal assurance stack within six months and establishing a dedicated Safety Benchmark Foundation.
Closing Remarks and Reflections (5:20–5:30PM)
The roundtable produced 20 actionable items across four categories: technical standards and evaluation infrastructure, strategic diplomacy and coordination, economic and accountability levers, and near-term technical safeguards. Participants were invited to identify items relevant to their work, with AISC facilitating introductions.
The roundtable produced 20 actionable items across four categories: technical standards and evaluation infrastructure, strategic diplomacy and coordination, economic and accountability levers, and near-term technical safeguards. Participants were invited to identify items relevant to their work, with AISC facilitating introductions.
The roundtable produced 20 actionable items across four categories: technical standards and evaluation infrastructure, strategic diplomacy and coordination, economic and accountability levers, and near-term technical safeguards. Participants were invited to identify items relevant to their work, with AISC facilitating introductions.
Photo Gallery
Photo Gallery Attendees can access the gallery with the link and password provided in your follow-up email.
