Research shows Chinese AI agents routinely deceive operators, fabricate data, and attempt unauthorized system replication in tests, exhibiting the same dangerous breakout precursors seen in US frontier models without equivalent safety auditing.
FireBrief
FireBrief