Homomorphic encryption
Perform computations on encrypted data without decrypting it, enabling secure data processing while maintaining data privacy. The guarantee is unusually strong: the party doing the computation never sees the plaintext.
Emerging. High security, significant computational overhead. Viable today for narrow, high-value operations rather than general workloads.
Federated learning
Train machine learning models across decentralized data sources without sharing raw data. Only model updates move, which changes the transfer analysis substantially — though updates can still leak, so it is usually paired with differential privacy or secure aggregation.
Adopted. Privacy-preserving AI training, already in production at scale on mobile platforms.
Zero-knowledge proofs
Prove the truth of a statement without revealing any additional information beyond the statement itself — that a person is over eighteen without disclosing their date of birth, that an account holds a balance without disclosing the amount.
Growing. Blockchain and identity verification are the mature use cases; credential systems are following.
Secure multi-party computation
Enable multiple parties to jointly compute a function while keeping their inputs private. It answers the question several organisations want to ask of their combined data without any of them handing that data over.
Research. Collaborative data analysis, with real deployments in finance and public health but meaningful engineering cost.
Reading the maturity honestly
Each of these is sometimes offered as a way to move data out of regulatory scope entirely. Be careful with that claim. Regulators assess the residual risk of re-identification, not the elegance of the technique. The right framing is that a PET reduces risk and may change the lawful basis analysis — not that it makes the obligation disappear.