Memory poisoning is a cybersecurity threat where an attacker deliberately corrupts or manipulates the stored memory or context files of an AI agent or software system. By injecting false or malicious data into memory, the attacker can alter the system’s behavior, cause it to make incorrect decisions, or even turn it malicious. This technique is particularly concerning in AI-driven environments, where memory is used to retain context, learn from interactions, or store sensitive information. Memory poisoning can lead to data breaches, loss of integrity, and compromised system security. Effective safeguards include memory integrity checks, regular audits, and robust access controls.