double-blind evaluation
AssessTechniques
An evaluation method where neither the model provider nor the evaluator can access the other side’s sensitive inputs in advance.
Why it's here
Placed in Assess: 2 article(s) of evidence from 2 source(s), led by research-stage coverage, with 1 in the last 30 days. Confidence 40%.
Evidence (2)
- 6Google DeepMind·9/23/2026researchGoogle DeepMind advances private AI compute with secure server-side memory
Google DeepMind is piloting a new approach to private AI compute that uses secure server-side memory to protect sensitive data during model interactions. The company also says it is testing what it describes as the world’s first double-blind AI evaluations as part of its safety and responsibility efforts.
- 7Google DeepMind·8/27/2026researchGoogle DeepMind pilots double-blind AI evaluations
Google DeepMind says it has piloted the world’s first double-blind evaluation of a proprietary frontier AI model. The approach uses Confidential Space in Google Cloud to keep both the model weights and external benchmark prompts private, with partners including the Singapore AI Safety Institute, OpenMined, AVERI, and MLCommons.