This is a reported privacy-verification approach for a deployed Gboard workload, but the available excerpt is insufficient to establish its security guarantees or broader national-security effect.
Limits of the evidence: Only a short press-source excerpt is supplied; it gives no linked technical paper, implementation details, or independent evaluation.
What to watch
Assessment revised Oct 8, 2026
Google Research Moves Federated Learning Into TEEs: Gboard Now Trains With Externally Verifiable Differential Privacy
MarkTechPostOct 4, 2026First report
First assessment, from MarkTechPost
No maturity ladder applies to this desk.
Sens.ai aggregates and assesses published reporting. The assessment above is machine generated; the original sources are authoritative.