Leaders from 31 countries are seeking closer international cooperation on the safety of advanced AI models, with Singapore proposing a system for reporting serious AI failures and near misses that is modeled on aviation safety practices.
Finnish President Alexander Stubb and Norwegian Prime Minister Jonas Gahr Støre convened a virtual meeting to discuss common safety standards, information sharing and potential international oversight mechanisms for frontier AI models – the most advanced AI systems.
The meeting followed a September declaration calling for mandatory safety testing, independent evaluations and greater transparency from AI developers.
Singapore Prime Minister Lawrence Wong proposed a voluntary international pilot under which participating countries would agree on categories of AI safety incidents and securely exchange information about failures and near misses.
Wong compared the approach to aviation, where reporting incidents – including those that do not cause actual harm – helps regulators and operators identify risks before accidents occur.
“Because when serious AI incidents happen, the lessons should not be confined to one country or one company,” Wong said in his official remarks. “We should be able to learn from what went wrong, identify the emerging risks, and take steps to prevent similar incidents happening elsewhere.”
The proposal would expand safety reporting beyond AI developers to include businesses, governments and other organizations using their models. Wong said users may encounter dangerous failures that fall outside existing reporting requirements.
In a joint statement, Finland and Norway said interested countries would continue discussions at the expert level to explore stronger AI safety measures and potential international mechanisms.
No binding international standards, reporting requirements or implementation timetable were announced.