Nobody Gets an A: Anthropic Tops FLI's 2026 AI Safety Index With a C+
TL;DR
FLI's Summer 2026 AI Safety Index: no company gets a B; Anthropic leads with a C+, xAI, DeepSeek and Mistral all fail.
The Future of Life Institute released the third AI Safety Index in July, grading nine major AI companies across six domains and 37 indicators. No company earned a B. Anthropic finished first with a C+ (2.66), followed by OpenAI at C (2.28) and Google DeepMind at C (2.01). Meta got a D+, China's Z.ai and Alibaba Cloud tied at D-, while Musk's xAI, DeepSeek and Mistral all received an F.
Anthropic led five of the six domains — transparency, safety framework, technical research and governance all came out ahead of peers — but reviewers added a sharp footnote. In February the company quietly dropped its pledge to never train a stronger model unless it could show its safety measures were adequate, and the panel flagged "questionable military engagements," including an indirect link to the Minab school strike.
The seven-member panel — Stuart Russell, David Krueger, Sharon Li, Tegan Maharaj, Sneha Revanur, Robert Trager and Yi Zeng — put every one of the nine companies at D+ or below on Existential Safety. Russell wrote: "Companies have backed away from earlier commitments to release new systems only with safety measures appropriate for their capability levels; now, they're planning to release them even if it's demonstrably unsafe to do so." Krueger was blunter: "AI companies' lack of progress towards credible AI Safety plans is scandalous. Even they are starting to get anxious as they race towards recursive self-improvement."
Between 2024 and 2026, Anthropic, OpenAI, Google DeepMind and Meta — all of which once banned military applications — walked those bans back and joined xAI and Mistral in chasing defense contracts. The C+ at the top of the class looks less like a passing grade than the tallest kid in a short row.
via Future of Life Institute / TIME / IEEE Spectrum
Anthropic led five of the six domains — transparency, safety framework, technical research and governance all came out ahead of peers — but reviewers added a sharp footnote. In February the company quietly dropped its pledge to never train a stronger model unless it could show its safety measures were adequate, and the panel flagged "questionable military engagements," including an indirect link to the Minab school strike.
The seven-member panel — Stuart Russell, David Krueger, Sharon Li, Tegan Maharaj, Sneha Revanur, Robert Trager and Yi Zeng — put every one of the nine companies at D+ or below on Existential Safety. Russell wrote: "Companies have backed away from earlier commitments to release new systems only with safety measures appropriate for their capability levels; now, they're planning to release them even if it's demonstrably unsafe to do so." Krueger was blunter: "AI companies' lack of progress towards credible AI Safety plans is scandalous. Even they are starting to get anxious as they race towards recursive self-improvement."
Between 2024 and 2026, Anthropic, OpenAI, Google DeepMind and Meta — all of which once banned military applications — walked those bans back and joined xAI and Mistral in chasing defense contracts. The C+ at the top of the class looks less like a passing grade than the tallest kid in a short row.
via Future of Life Institute / TIME / IEEE Spectrum
