FLI Releases AI Safety Index: No Global Model Scores Above C+

The Future of Life Institute (FLI) has published its Summer 2026 edition of the AI Safety Index, a comprehensive assessment of safety levels across major AI models worldwide. The results show that no evaluated model achieved a grade higher than C+.

FLI is a prominent research organization focused on mitigating extreme risks from transformative technologies including AI, biotechnology, and nuclear weapons. Its AI Safety Index aims to provide an independent, quantifiable standard for safety evaluation, covering dimensions such as model transparency, robustness testing, and alignment research progress.

According to the report's findings circulating via Chinese tech news, the assessment covered models from multiple leading AI companies and research institutions. Industry analysts had been anticipating the report as an important reference point for safety governance.

The rating arrives amid intensifying global focus on AI safety. Governments and international bodies are accelerating AI governance frameworks, including the implementation of the EU AI Act and deepening UN-level discussions. FLI's report offers policymakers a third-party assessment independent of corporate self-reporting.

A C+ rating does not indicate direct harm from these models, FLI noted. Rather, it reflects composite performance across safety dimensions including interpretability, adversarial robustness, and value alignment — areas where some progress has been made alongside clear shortcomings.

FLI stated it will continue to update the AI Safety Index and urged AI developers to embed safety evaluation as a core component of model training and deployment. Industry observers expect such external safety assessments to become standard practice across the AI field.