Singapore’s prime minister wants governments to exchange information about serious AI failures and near misses, borrowing an idea from an industry that has spent decades studying accidents: aviation.
Lawrence Wong proposed a voluntary international pilot on October 28 at a virtual meeting on frontier AI safety hosted by Norway and Finland. Participating countries would first agree on what qualifies as an incident and what information could be shared securely.
The proposal addresses a conspicuous weakness in AI oversight. Much of what the public knows about dangerous model behavior comes from the companies building the systems. Businesses, public agencies and other users may encounter failures that never become part of a broader record.
Wong argued that those encounters matter even when no one is hurt. A near miss can expose a flaw before a more serious incident occurs. Aviation investigators have long used that principle to improve safety procedures.
No participating countries, reporting requirements or implementation dates were announced. Questions about commercial confidentiality, national security and the willingness of organizations to disclose failures remain open.
The plan is deliberately modest. Wong urged governments not to begin with an elaborate new international institution, but with a small system they could test. That may make cooperation easier at a time when countries remain divided over how far AI regulation should go.
Whether the pilot amounts to more than a diplomatic gesture will depend on what governments agree to report, and whether those reports contain enough detail for others to learn from them.
Singapore Prime Minister Lawrence Wong in May 2025. Photo: U.S. Navy Petty Officer 1st Class Alexander Kubitza / U.S. Department of Defense (public domain).
Source: Prime Minister’s Office of Singapore, October 28, 2026.
