by Future of Life Institute·Latest: Winter 2025·Source ↗
Letter-grade ratings of frontier AI labs across six safety domains: risk assessment, current harms, safety frameworks, existential safety, governance, and information sharing.
Rows are organizations; columns are the dimensions FLI Index scores. Cell text is the published grade (overall first, then per-pillar).
| Organization | Overall | Current Harms | Existential Safety | Governance & Accountability | Information Sharing | Risk Assessment | Safety Frameworks |
|---|---|---|---|---|---|---|---|
| Alibaba Cloud | D- | D+ | F | D+ | D+ | D | F |
| Anthropic | C+ | C+ | D | B- | A- | B | C+ |
| DeepSeek | D | D+ | F | D | C- | D | F |
| Google DeepMind | C | C | D | C- | C | C+ | C+ |
| Meta AI (FAIR) | D | D+ | F | D | D- | D | D+ |
| OpenAI | C+ | C- | D | C+ | B | B | C+ |
| xAI | D | F | F | D | C | D | D+ |
| Z.ai | D | D | F | D | C- | D+ | D- |
| Z.ai | D | D | F | D | C- | D+ | D- |
| Wave | Published | Orgs | License | Links |
|---|---|---|---|---|
| Winter 2025latest | 2025-12-02 | 8 | fair-use-citation | |
| Summer 2025 | 2025-07-17 | 7 | fair-use-citation | |
| December 2024 | 2024-12-11 | 6 | fair-use-citation |
Letter-grade ratings of frontier AI labs across six safety domains: risk assessment, current harms, safety frameworks, existential safety, governance, and information sharing.
Evidence collected up until November 8, 2025. Survey responses submitted by Anthropic, OpenAI, Google DeepMind, xAI, and Z.ai. Meta, DeepSeek, and Alibaba Cloud did not submit survey responses. The 'Survey Responses' dimension is not a scored domain but tracks participation.
Grades are mirrored from upstream sources. We do not score organizations ourselves — see the source link for full methodology and per-indicator detail. Citation of published grades is consistent with fair-use attribution.