The Future of Life Institute’s latest AI Safety Index found that even the industry’s highest-rated AI developers fell well short of top marks, with three companies receiving failing grades.
The Future of Life Institute (FLI) released its Summer 2026 AI Safety Index. The non-profit focused on reducing risks from advanced technologies said that no leading AI developer demonstrated safety practices worthy of a grade higher than C+.
The report evaluated nine of the world’s leading AI companies across 37 indicators spanning six areas: risk assessment, current harms, safety frameworks, governance, and information sharing.
The assessment combines publicly available evidence with company survey responses, both reviewed by an independent panel of AI researchers and governance experts.
Key Takeaways
- No company earned an overall grade better than a C+.
- Anthropic ranked first, followed by OpenAI and Google DeepMind.
- Three companies, xAI, DeepSeek, and Mistral, received failing grades.
- Meta was the most improved company, while xAI recorded the largest decline since the previous index.
- Reviewers said AI companies expanded their written safety policies. However, many still leave key questions unanswered, including what level of risk should delay a model’s release, whether independent experts should review that decision, and who has the authority to stop a release.
The reviewers said the low scores reflected persistent weaknesses in how companies manage AI safety. Even though many of these frontier AI developers published more detailed policies and made new public commitments over the past year, many companies still do not clearly define when AI models should be delayed, who can make that decision, or what independent oversight is required before release.
The rankings reflect the findings. Anthropic received the highest overall score with a C+. U.S.-based OpenAI and Google DeepMind each earned a C. Meta ranked fourth with a D+, while Z.ai and Alibaba Cloud each received a D-. Three companies, xAI, DeepSeek, and Mistral, received failing grades.
The report also tracked how each company’s rank changed since the previous edition. Meta was the biggest improver, climbing from sixth to fourth place, while xAI recorded the largest decline, falling from fourth to seventh. Although Anthropic retained the highest overall ranking, OpenAI moved into first place in the report’s Risk Assessment category because of its broader safety evaluations and greater use of independent external testing.
