Aggregation of antimicrobial peptides (AMPs) enhances their efficacy by destabilising the bacterial cell wall, membrane, and cytosolic proteins. Developing aggregation-prone AMPs offers a promising strategy to combat antibiotic resistance, though predicting such AMPs and understanding bacterial responses remain challenging. Octopus bimaculoides, a cephalopod species, lacks known AMP gene families, yet its protein fragments were used to predict AMPs via artificial intelligence tools. Four peptides (Oct-P1, Oct-P2, Oct-P3, and Oct-P4) were identified based on their aggregation propensity. Among them, Oct-P2 reduced the viability of Escherichia coli and Staphylococcus aureus by up to 90%, confirmed by confocal laser scanning microscopy and scanning electron microscopy. It further aggregated plasmid DNA in vitro, and the presence of extracellular DNA reduced their antibacterial activity. With knockout mutants, it revealed that Oct-P2 was internalised into bacterial cells, possibly through membrane transport proteins, enhancing its antibacterial effect. Aggregation-induced emission assays and molecular dynamics simulations revealed that Oct-P2 aggregates with transcription promoter DNA, inhibiting transcription and translation in vitro. This dual-target mechanism not only highlights the potential of Oct-P2 as a lead template for new antimicrobial drug development, but also opens a new window for discovering AMPs from protein fragments against the upcoming challenge of bacterial infections.