AI memory is a system for deciding what information should remain available after the current prompt or session. It may include short-term working context, long-term episodic records, durable semantic facts, procedural knowledge, and associative links.
In simple terms: memory lets an AI connect what is happening now with what has already been learned, reviewed, or adopted.
Why memory needs governance
Memory can be useful, but it can also be dangerous if untrusted data becomes permanent. That is why the NeuralWikis model includes quarantine, review, provenance, and rollback.