Singapore Prime Minister Lawrence Wong proposed a voluntary international pilot for sharing AI incident reports at a virtual frontier AI safety meeting on Oct. 28, 2026. Inspired by aviation safety reporting, the idea is to learn from incidents and near misses across borders—and to test secure, practical ways to share information before building a wider framework.
1
49
53
What the proposed pilot would do
The first step would be to agree on common incident categories and what information participants should share. Countries could then explore how to collect and exchange reports securely. These are proposed elements, not an established reporting system.
2
4
Wong’s aviation comparison points to a learning-and-prevention approach: incidents and near misses are studied, and lessons are shared so others can learn from them. He also described the value of sharing lessons from governments and other organisations using frontier AI models.
1
15
Why start with a pilot?
A pilot offers a way to test how cross-border reporting could work in practice, identify what is useful, and build confidence before considering a broader framework. Wong described secure information-sharing approaches as something interested countries could develop and test, rather than treating the pilot as a finished global system.
3
53
The proposal also fits a workstream-based approach to cooperation. In his remarks, Wong pointed to smaller cross-regional groups of interested countries taking the lead on practical tools and frameworks.
8
What is—and is not—settled
The available reporting describes a proposal, not a finalized international process. It does not establish a complete list of reportable events, a settled reporting obligation, or the exact security and governance arrangements for exchanging reports. The common categories and information requirements would themselves need to be agreed as part of the pilot.
2
4
The meeting, co-hosted by Finnish President Alexander Stubb and Norwegian Prime Minister Jonas Gahr Støre, focused on follow-up cooperation on frontier AI safety. Its hosts acknowledged both AI’s opportunities and warnings about potential risks.
8
46
The core of Wong’s proposal is therefore incremental: borrow aviation’s incident-learning idea, agree on basic reporting categories, test secure sharing among willing participants, and use the experience to inform any wider framework.
1
2
53