The UK will host a convention in San Francisco for discussions with AI builders on how they’ll put into observe commitments made on the AI Motion Summit.
Held on 21 and 22 November, the occasion will characteristic workshops and discussions targeted on AI security forward of France internet hosting the AI Motion Summit in February 2025.
Earlier this yr, 16 firms from around the globe, together with these from the US, EU, Republic of Korea, China, and the UAE, agreed to publish their newest AI security frameworks ahead of the next Summit.
These frameworks will lay out their plans to deal with probably the most extreme potential AI dangers, together with if unhealthy actors misuse the expertise.
As a part of these commitments, firms additionally agreed to cease the deployment or improvement of any fashions if their potential dangers can’t be sufficiently addressed.
Sharing AI security frameworks
The occasion shall be a second for AI firms to take inventory and share concepts and insights to assist the event of their AI security frameworks by way of a focused day of talks between signatory firms and researchers.
Science, Innovation and Know-how Secretary Peter Kyle defined: “The convention is a transparent signal of the UK’s ambition to additional the shared international mission to design sensible and efficient approaches to AI security.
“We’re simply months away from the AI Motion Summit, and the discussions in San Francisco will give firms a transparent give attention to the place and the way they’ll bolster their AI security plans constructing on the commitments they made on the final Summit in Seoul.”
From at this time, attendees are additionally urged to share ideas on potential areas of debate at November’s convention, together with present and present proposals for developer security plans, the way forward for AI mannequin security evaluations, transparency and strategies for setting out totally different danger thresholds.
A community of AI institutes
Discussions, co-hosted with the Centre for the Governance of AI and led by the UK’s AI Security Institute (AISI), will assist construct a deeper understanding of how the Frontier AI Security Commitments are being carried out.
The UK’s AI Security Institute is the world’s first state-backed physique devoted to AI security. The UK has continued to play a world management position in growing the rising worldwide community of AI Security Institutes, together with its landmark settlement with the US earlier this yr.
The convention has been designed as a discussion board for attendees to change concepts on greatest practices in implementing the commitments.
This ensures a clear and collaborative method for builders as they refine their AI security frameworks forward of the AI Motion Summit.
It follows the US authorities’s announcement yesterday of the primary assembly of the Worldwide Community of AI Security Institutes, which is able to happen on the times earlier than 20-21 November 2024 in San Francisco.
The UK launched the world’s first AI Security Institute at Bletchley Park final November, and since then, nations around the globe have raced to determine their very own AI security testing our bodies.
The US-hosted convening will carry collectively technical specialists on synthetic intelligence from every nation’s AI security institute or equal government-backed scientific workplace to align on the community’s precedence work areas and start advancing international collaboration and data sharing on AI security.
