Google DeepMind Launches World's First Double-Blind AI Evaluation Using Cryptographic Technology to Ensure Testing Integrity

Google DeepMind launches the world's first double-blind AI evaluation using cryptographic technology, partnering with global safety organizations to test its Gemini Flash Lite model against confidential benchmarks in a secure environment where neither party can access the other's sensitive data, setting a new standard for trustworthy AI oversight.

• Google DeepMind launches the world's first double-blind evaluation of a proprietary frontier AI model, using cryptographic technology to prevent benchmark contamination and ensure evaluation integrity.
• The pilot partners with the Singapore AI Safety Institute, OpenMined, AVERI, and MLCommons, testing a Gemini Flash Lite model against confidential benchmarks inside a secure cryptographic environment where neither the evaluator can see model weights nor Google can see test prompts.
• The breakthrough eliminates a longstanding tradeoff in AI testing by allowing independent organizations to rigorously assess advanced models without compromising intellectual property or data sovereignty, setting a new standard for trustworthy AI oversight.