The UK will host a conference in San Francisco for discussions with AI developers on how they can put into practice commitments made at the AI Seoul Summit.

To be held across the 21 and 22 November, the event will feature a number of workshops and discussions focused on AI safety ahead of France hosting the AI Action Summit in February 2025.

Earlier this year, 16 companies from across the globe including those from the US, EU, Republic of Korea, China and the UAE, agreed to publish their latest AI safety frameworks ahead of the next Summit.

These frameworks will lay out their plans to tackle the most severe potential AI risks, including if the technology was misused by bad actors. As part of these commitments, companies also agreed to stop the deployment or development of any models if their potential risks cannot be sufficiently addressed.

The event will be a moment for AI companies to take stock and share ideas and insights to support the development of their AI safety frameworks through a targeted day of talks between signatory companies and researchers.

Science, Innovation and Technology Secretary Peter Kyle said:  

The conference is a clear sign of the UK’s ambition to further the shared global mission to design practical and effective approaches to AI safety.

We’re just months away from the AI Action Summit, and the discussions in San Francisco will give companies a clear focus on where and how they can bolster their AI safety plans building on the commitments they made in Seoul.

From today, attendees are also urged to share thoughts on potential areas of discussion at November’s conference, including existing and current proposals for developer safety plans, the future of AI model safety evaluations, transparency and methods for setting out different risk thresholds.

Co-hosted with the Centre for the Governance of AI and led by the UK’s AI Safety Institute (AISI), discussions will help build a deeper understanding of how the Frontier AI Safety Commitments are being put into practice.

The UK’s AI Safety Institute is the world’s first state-backed body dedicated to AI safety, and the UK has continued to play a global leadership role in developing the growing international network of AI Safety Institutes – including its landmark agreement with the US earlier this year.

The conference has been designed as a forum for attendees to exchange ideas on best practice in implementing the commitments, ensuring a transparent and collaborative approach for developers as they refine their AI safety frameworks ahead of the AI Action Summit.

It follows the US government yesterday announcing the first meeting of the International Network of AI Safety Institutes, which will take place in the days before from 20-21 November 2024, in San Francisco. The UK launched the world’s first AI Safety Institute at Bletchley Park last November, and since then nations around the world have raced to establish their own AI safety testing bodies.

The convening hosted by the US will bring together technical experts on artificial intelligence from each country’s AI safety institute, or equivalent government-backed scientific office, in order to align on priority work areas for the Network and begin advancing global collaboration and knowledge sharing on AI safety.