BETA RELEASE

FLI AI Safety Index: Summer 2025

Tertiary report Future of Life Institute
Published: 2025-07-01 Accessed: 2026-05-09
Original source ↗ Archived copy ↗

Summary

The Summer 2025 FLI AI Safety Index evaluates safety practices of seven leading AI companies across six domains, finding modest progress but major gaps, especially in control risk.

Key quotes

"The Control-Problem remains unsolved: Despite their explicit ambitions to develop artificial general intelligence (AGI) ... the review panel again found the current strategies of all companies inadequate for ensuring that these systems remain safe and under human control."
"Grading: Uses the US GPA system for grade boundaries: A+, A, A-, B+, [...], F letter values corresponding to numerical values 4.3, 4.0, 3.7, 3.3, [...], 0."

The two‑page index provides an executive summary, methodology, and a list of the seven evaluated companies (Anthropic, OpenAI, DeepMind, x.AI, Meta, Zhipu AI, Deepseek) and their grades across 34 safety indicators. It is authored by an independent panel of AI safety experts.