Nobody Gets an A: The 2026 AI Safety Index Grades the Top Labs — Anthropic Leads With a C+

The report card is in, and the whole class is struggling. The Future of Life Institute released its Summer 2026 AI Safety Index this month, grading nine of the world's leading AI developers on how seriously they take safety — and no company earned an A or a B. The top of the class, Anthropic, managed only a C+; the message from the independent review panel is that even the industry's safety leader is doing barely-passing work.

The full ranking is sobering. Anthropic took first with a C+, OpenAI slipped from a C+ last time to a C, and Google DeepMind came third, also at C. Meta earned a D+, while Z.ai and Alibaba Cloud both scored a D-. At the bottom, three labs failed outright: xAI, DeepSeek, and France's Mistral all received an F. As the institute pointed out, those failing grades come one each from the United States, China, and Europe — a sign that weak safety practice is not confined to any single region or regulatory regime.

The index is not a vibe check. A panel of independent AI experts scored each company across domains including risk assessment, handling of current harms, safety frameworks, existential-safety planning, governance and accountability, and information sharing. The exercise is deliberately comparative: it grades labs against best practices the field itself has articulated, which is why so many land at C or below even as they ship increasingly capable models. When the people building frontier systems can't clear their own bar, the gap between capability and control becomes the story.

Two trends alarmed the reviewers in particular. First, militarization: between 2024 and 2026, companies that once explicitly banned military applications — Anthropic, OpenAI, Google DeepMind, and Meta among them — quietly reversed course and began courting defense partnerships, joining xAI and Mistral. Second, the erosion of commitments: several of the same labs have weakened or voided earlier pledges to pause development unilaterally if their models approached dangerous "redlines." Safety promises made in calmer times are being renegotiated under competitive pressure.

The context makes the grades hit harder. This is the same industry racing toward trillion-dollar valuations and public listings — Anthropic alone is running near a $47 billion revenue pace and eyeing an October IPO — while national governments from Washington to Seoul pour hundreds of billions into scaling it up. The Safety Index is a reminder that the money and capability curves are climbing far faster than the safety curve, and that "best in class" currently means a C+. For regulators now drafting pre-market evaluation regimes, and for a public increasingly living alongside these systems, that is precisely the point the Future of Life Institute wants on the record.