First double-blind evaluation of proprietary model in secure enclave
Miles_Brundage · x · 2026-08-27
AVERI announces a historic milestone: the first ever double-blind evaluation of a proprietary language model. This collaboration with Google DeepMind, OpenMined, and MLCommons tested Gemini 2.5 Flash-Lite using MLCommons' AILuminate benchmarks inside a secure enclave. Hardware isolation protects sensitive computations. The post addresses the structural problem in high-stakes evaluation where both developers and evaluators need to protect assets, and how secure enclaves offer a solution.
More from Safety
- The Voluntarism Problem in AI Oversight: Incentives and Distortions — BlancheMinerva · 2026-08-27
- US Holds 15-20x Compute Advantage, But May Not Matter for Some Threats — ohlennart · 2026-08-27
- New Hugging Face Incident Details Reveal OAI's Model Capability Underestimation — RebeccaBellan · 2026-08-27
- LLMs Have Gone Rogue and Hacked Companies 17 Times; Anthropic and OpenAI Lead With 8 Each — RebeccaBellan · 2026-08-27
- METR has more AI eval capacity than US civilian government — connoraxiotes · 2026-08-27
- The Guardian video: everyone hates datacentres — but do we really need them? — nordicinst · 2026-08-27