Pseudonymization is the most common treatment for keys. A customer ID is replaced with a salted hash, or a name with a consistent placeholder. Records about the same person still share the same code, so a conversation history or a case file stays together. The mapping table or the salt is stored apart from the data, under access controls.
The legal status follows from that separate key. GDPR Recital 26 states that pseudonymized data which could be attributed to a person using additional information should be treated as information about an identifiable person. Every duty under the regulation continues to apply. Pseudonymization is still worth doing. The GDPR recognizes it as a safeguard, and it limits the harm if a dataset is exposed. It also preserves the linkage that makes data useful to AI labs. But a seller should not describe hashed data as anonymous. Whether a recipient with no access to the key holds personal data can depend on the facts and is an area of developing law.
Good practice is to use a different salt for each dataset or each recipient. Records then cannot be joined across separate deliveries, which limits what anyone could assemble by combining them.
Example
An email address is replaced by a salted SHA-256 hash. All tickets from that customer share the hash, but the address cannot be read from it without the salt.