Thursday, July 16, 2026

Anthropic tops AI safety index but still earns only a C+

AnthropicOpenAI

The Future of Life Institute's Summer 2026 AI Safety Index, which grades nine leading AI developers across 37 indicators in six domains, ranked Anthropic first with an overall grade of just C+, followed by OpenAI and Google DeepMind at C, Meta at D+, and xAI, DeepSeek and Mistral effectively failing. The expert review panel, which includes UC Berkeley's Stuart Russell, said even the most safety-focused labs are performing mediocrely against their own stated standards as their models are wired into cybersecurity, healthcare and autonomous agents. It also flagged a consistent pattern of frontier labs quietly weakening or dropping safety commitments made during fundraising as competitive pressure mounts.