A peer-reviewed paper published on 30 September describes detectable marks in amino-acid sequences and digital structure models. For binders tested against three targets, it found no significant deterioration in binding across the compared groups. This does not validate every protein function or constitute a treatment in humans.

A subsequent computational modification could also remove the structure watermark. The authors work for Alphabet, which funded the research; independent experimental replication is not established here. The signature should not be confused with irremovable protection or certification of harmlessness.

In practice, such a trace might resemble a batch number that travels with a design. It could help a laboratory determine whether a sequence came from a particular tool and review provenance in a research database. It would not, by itself, show that a protein is correctly described, effective or safe.

Before the idea becomes routine infrastructure, other protein types, modifications and false alarms across many checks need testing. Rules for managing detection keys and agreement between laboratories on what a positive result means are also important. A batch number helps traceability; it does not replace quality testing.

An optimistic scenario is a limited pilot involving several cooperating laboratories within six to eighteen months. This is an editorial estimate conditional on further validation; there is no defensible date for broad biosecurity screening. The illustration is schematic and does not depict an experimentally measured molecule.