memorisation

A communication question with a technical core. The answer that scores uses one good analogy, stays accurate, and ends with options and a decision.

Fine-tuning moves data from a database you can query and delete into weights you cannot. The scored answer covers memorisation, tenant separation, poisoning and the deletion problem.

Language models are trained to predict text, and sometimes the best prediction is the exact text they saw. Rare, repeated and sensitive strings are the risk.

The leak is in the weights, so the playbook differs from a database breach. The scored answer confirms, contains at the output, scopes from data lineage, and decides between filtering, retraining and withdrawal.

A research-depth question. The scored answer explains why 'was this in training' blurs at web scale, why many published attacks measured distribution shift instead, and what questions remain answerable.

Generic leakage benchmarks will not tell you whether your customers' records got easier to extract. Planted canaries, matched controls and an attack-economics view of severity will.