A safety evaluation of nine major global AI companies, released on the 7th by the U.S.-based nonprofit Future of Life Institute (FLI), found that the highest grade was barely a 'C+', with not a single company earning an A or B. The '2026 First-Half AI Safety Index' targeted Anthropic, OpenAI, Google DeepMind, Meta, Z.ai, Alibaba Cloud, xAI, DeepSeek, and Mistral. An independent expert panel of seven assessed them across six categories. Anthropic placed first overall with a 'C+', followed by OpenAI and Google DeepMind each with a 'C', Meta with a 'D+', Z.ai and Alibaba Cloud with a 'D-', and xAI, DeepSeek, and Mistral with an 'F'. The FLI strongly criticized companies for breaking their own self-imposed 'red line' promises, noting that Anthropic withdrew its previous commitment in February not to train AI systems unless it could guarantee sufficient safety measures. FLI co-founder Max Tegmark remarked, "AI companies are warning about the dangers of artificial superintelligence (ASI) themselves, yet they are not stopping the race to create it."