More than 20 countries warn about advanced AI risks and call for controls

The signatories called on companies to adopt transparent security protocols, including mandatory testing and independent assessments.