Neuro-Alignment Gauge: A benchmark for cognitively grounded, interpretable brain-decoding foundation models

We are building the Neuro-Alignment Gauge to verify AI safety. By training a multimodal brain-decoding foundation model as biological ground truth, we use neural representation geometry to measure the ‘cognitive distance’ between AI and human imagination. This creates a rigorous safety testbed for cognitively grounded AI.

Biography

How do we quantify the distance between AI representations and the brain’s visual imagery space? Ninon Lizé Masclef is a Brain & AI researcher and Foresight Fellow 2026. She co-leads research on neural decoding of visual perception and imagery at MIT Media Lab. Her first brain-computer interface was named one of France’s top 100 inventions of 2024, and she is an O’Shaughnessy Fellow 2025.