2026 인공지능 안전 서울 포럼

프로그램

프로그램

The International Network Workshop in Seoul

9^th July 2026, IT Square, Coex Center Seoul

Agenda

08:00~09:00	Registration and coffee reception
09:00~09:30	Traditional performance & opening remarks "Geulssedang," a traditional performing arts group specializing in calligraphy, to open the event. Opening remarks to follow from the Korea AI Safety Institute, Korea’s Ministry for Science and ICT and the UK’s AI Security Institute.
	[Session 1] Shared tools in AI evaluations
09:30~11:00	Main presentation UK AISI’s ‘engineering playbook’ Short showcases UK AISI’s evaluation database UK AISI’s Inspect Scout EU AIO’s risk modelling work France INRIA’s PANAME project Breakout groups Following presentations, presenters will join breakout groups (7-10 individuals) to facilitate Q&A around the topics covered and brainstorm what further shared tools could support AI evaluations.
11:00~11:30	Coffee break
	[Session 2] Best practice in agentic evaluations
11:30~13:00	Main presentations UK AISI on inference scaling UK AISI on open-world evaluations Breakout groups Following presentations, facilitated breakout groups will discuss challenges and solutions facing agentic evaluations around the following themes: Inference scaling Scaffolding, including specialised tooling Agentic security evaluations Next generation / open-world evaluations
13:00~14:00	Lunch Attendees external to the International Network to leave after lunch.
	[Session 3] Next steps in Network best practices
14:00~15:30	Network members to discuss open research questions around tools and best practices needed in tackling future agentic evaluations.
15:30~15:45	Closing remarks K-AISI and UK AISI to provide closing remarks.