Chosen plaintext is data that an attacker can deliberately select and send to a cryptographic system for encryption. This creates a chosen-plaintext attack (CPA) setting, where the adversary controls inputs instead of only observing traffic. It is a standard model in cryptography because many real systems expose encryption APIs, file upload features, or protocol messages that can be influenced by an attacker.
It matters because controllable inputs can reveal patterns, weaknesses, or hidden structure in a cipher or mode of operation. Researchers use chosen plaintexts to test whether an algorithm resists adaptive probing, to measure security margins, or to build proofs of semantic security. In practice, an attack may require very large amounts of chosen plaintext, which is why a result can be important in the lab but still impractical in the field. Defenders reduce exposure by limiting encryption oracles, using authenticated encryption, and ensuring implementations do not leak useful feedback.