Fisheries worldwide are increasingly adopting on-vessel cameras, electronic monitoring (EM), for catch recording. AI-driven analysis of EM data offers the potential to improve data quality by providing accurate counts of target and non-target species for species-specific assessments. This presentation will share initial results and discuss challenges from a new project applying AI to EM in Australia's sub-Antarctic fisheries that target Patagonian toothfish (Dissostichus eleginoides), including those around Heard Island and McDonald Islands (HIMI) and Macquarie Island. The HIMI fishery is a Marine Stewardship Council certified longline fishery that also incidentally catches grenadiers (Macrouridae), deepwater skates (Bathyraja spp.), and to a lesser degree, morid cods (Antimora rostrata). The goal is to test AI’s feasibility in delivering accurate counts and measurements of both target and non-target species, with new technologies enabling near real-time catch analysis. Automated species identification using AI could significantly reduce costs for industry and regulatory agencies while enhancing fishery monitoring coverage.