Researchers developed a deep‑learning framework that can reprogram protospacer‑adjacent motif (PAM) recognition for CRISPR–Cas enzymes, enabling tailored targeting specificities. The model transforms PAM recognition constraints that traditionally limit where Cas nucleases can cut, offering a computational path to expand editable loci without engineering new protein scaffolds. Authors propose this AI approach as a tool to accelerate CRISPR targeting design for research and therapeutic programs, enabling teams to consider previously inaccessible genomic sites in development plans.