At the very beginning of the new decade, the COVID-19 pandemic has badly hit modern human societies. SARS-CoV-2, the causative agent of COVID-19 carries dozens of new mutations in its genome. Herein, we made an effort to find new antiviral peptides (AVPs) against SARS-CoV-2. Gladly, with the help of Machine Learning algorithms, and Supported Vector Machine, we have invented three new AVPs against the SARS-CoV-2. Antiviral peptides viz., Seq12, Seq12m, and Seq13m can block the receptor binding domain (RBD) of the SARS-CoV-2, necessary for communication with the angiotensin-converting enzyme 2 (ACE2). In addition, these AVPs retain their antiviral properties, even after the insertion of dozens of new mutations (Rosetta, and FoldX based) in the RBD. Further, Seq12, and Seq12m showed negligible cytotoxicity. Besides, the binding free energy calculated using MM-PB/GBSA method is also in agreement with the molecular docking studies performed using HADDOCK. Furthermore, the molecular interactions between AVPs and the viral membrane protein (M) also showed a thermodynamically favorable interaction, suggesting it could eventually inhibit the viral re-packaging process. In conclusion, this study suggests AVPs viz., Seq12, Seq12m, and Seq13m embrace importance as a potential anti-SARS-CoV-2 therapeutic. These AVPs could also aid virus diagnostic tools in the future.