Boron　nitride　nanosheet　(BNNS)　reveals　a　huge　potential　in　preparing　highly　flame-retardant　polymer-based　thermal　conductive　composite,　but　is　limited　by　its　difficult　exfoliation　and　functionalization.　Here,　hexagonal　boron　nitride　(hBN)　was　simultaneously　exfoliated　and　flame-retardant　functionalized　into　BNNS　via　one-step　ball　milling　process　based　on　the　synergetic　effect　of　mechanical　shear　and　chemical　peeling　of　ammonium　phosphate　and　sodium　hydroxide.　Then　the　epoxy　(EP)-based　composites　containing　hBN　or　BNNS　were　prepared　by　solution　blending　and　program-controlled　curing.　The　possible　mechanochemical　reaction　mechanisms　were　proposed　according　to　the　incorporation　of　density　functional　theory　(DFT)　calculations　and　chemical　structure　characteristics.　As　one　of　potential　applications,　the　obtained　flame-retardant　functionalized　BNNS　(BNNS1　and　BNNS2)　were　used　as　multifunctional　additives　for　fabricating　high-performance　EP-based　thermal　conductive　composites　with　excellent　flame　retardancy.　As　expected,　the　obtained　EP-based　composites　containing　only　5　wt%　BNNS　exhibited　a　superior　flame　retardancy　with　a　dramatic　decrease　in　the　values　of　peak　heat　release　rate　(PHRR),　the　total　heat　release　(THR),　the　smoke　production　rate　(SPR)　and　the　total　smoke　production　(TSP)　corresponding　to　60.9%,　35.7%,　44.3%　and　38.8%　reductions,　respectively,　compared　to　neat　EP.　The　dramatical　enhancement　in　flame　retardancy　was　mainly　attributed　to　the　catalytic　charring　effect　and　physical　barrier　action　of　flame-retardant　functionalized　BNNS,　led　to　the　formation　of　a　compact　and　robust　char　layer　during　combustion　to　protect　the　underlying　polymer.　Simultaneously,　due　to　uniform　dispersion　and　strong　interfacial　adhesion,　the　incorporation　of　BNNS　not　only　increased　the　thermal　conduction　paths　by　increasing　specific　surface　area,　but　also　reduced　the　interfacial　thermal　resistance　(Rb)　caused　by　phonon　scattering,　leading　to　an　enhancement　(312.4%　and　397.0%)　in　the　TC　of　EP/BNNS　composites　at　30　wt%　BNNS1　and　BNNS2,　respectively.　©　2020　Elsevier　B.V.