Someone finally graded the AI labs on safety and the best in the class got a C+
The Future of Life Institute published its Summer 2026 AI Safety Index, scoring nine leading AI companies - including Anthropic, OpenAI, Google DeepMind, xAI, Meta, DeepSeek, Alibaba and Mistral - across 37 indicators in six areas, from risk assessment to governance, judged by an independent panel of seven researchers. Anthropic topped the table with a C+. OpenAI and Google DeepMind landed at C, Meta at D+, and xAI, DeepSeek and Mistral effectively failed. The panel's blunt finding: several labs have quietly walked back safety commitments they made while raising money. It's an outside scorecard, not a lab grading its own homework.
What it actually means
For a month the only question here has been price - swap to the cheaper model, it's a config line. This is the column nobody puts in the swap tutorial. When you build on a model you're not just buying tokens; you're inheriting a company's judgement about how it behaves when it matters - and now there's an independent, dated read on whose judgement that is. The best grade was a C+, so this isn't 'use these, avoid those'; everyone ships on a C+ somewhere. It means governance belongs next to benchmark and cost when you pick a foundation - heavier for anything regulated, anything touching customer data, anything you can't easily unwind. The cheap model stays cheap. The question is whose promises you're standing on when you build on it.
You're not just buying tokens - you're building on a company's judgement, and the best grade in the class was a C+.
Source: Future of Life Institute · Thursday 16 July 2026