Future of Life Institute Summer 2026 AI Safety Index: Anthropic C+ (top), OpenAI C (down from C+), Google DeepMind C. Meta improved D to D+ (6th to 4th). xAI fell to F after SpaceXAI merger — joins DeepSeek and Mistral as failing grades across US, China, Europe.
Panel notes weakened pause commitments and military AI pivots at leading labs. Max Tegmark: real “race to the top” requires regulation; cites EU AI Act and US risk-conscious shift.
Evidence collected through June 3, 2026.