No frontier AI lab cleared a C+ in the Future of Life Institute’s Summer 2026 AI Safety Index, published July 7, with Anthropic topping the nine-company field at C+, OpenAI and Google DeepMind at C, Meta at D+, and xAI, DeepSeek and Mistral pulling outright Fs. The report, built from 37 indicators across six domains and adjudicated by seven outside reviewers drawn from UC Berkeley, Oxford, the University of Montreal, University of Wisconsin-Madison and HEC Montréal, uses an evidence cutoff of June 3. TIME called the grades “unforgiving.” That reads like restraint.
The more revealing story sits underneath the letter grades. Reviewers noted that Anthropic, OpenAI, Google DeepMind and Meta have all weakened or voided earlier commitments to pause development unilaterally if their systems approached specified danger thresholds, with some pledges now made contingent on what competitors do. The panel called this “moving the goalpost” and said it has “undermined safety frameworks across the board.” Between 2024 and 2026, those same four labs also reversed earlier bans on military applications, joining xAI and Mistral in actively pursuing defense partnerships. Anthropic, the class valedictorian, was singled out for “questionable military engagements,” including a reported link to the Minab school strike that FLI says caused mass civilian deaths. Alibaba Cloud and Z.ai deny U.S. allegations of military ties.
Motion within the rankings tells its own story. Meta climbed to fourth from sixth. xAI fell to seventh from fourth. OpenAI slipped from C+ to C. Anthropic led five of six domains; OpenAI took the top slot in Risk Assessment.
Stuart Russell, the Berkeley computer scientist who served as a reviewer, said “the capabilities race has become more extreme,” with firms that “are planning to release [new systems] even if it’s demonstrably unsafe to do so.” David Krueger of the University of Montreal called the lack of progress toward credible safety plans “scandalous,” adding that CEOs are “still not telling people how urgent the risk is and how unprepared they are.”
FLI chair Max Tegmark was blunter: “AI companies are sprinting toward a cliff. Despite acknowledging the great risks of artificial superintelligence, they continue racing to build it.” He also claimed the survey is “already having an impact in internal discussions,” and five of the nine companies filled it out this cycle, up from near-zero participation in the first edition.
That last detail is the one worth sitting with. A voluntary safety scorecard that started as an exercise in shouting into a void now has majority participation from the labs it grades, and the labs are still walking back their own pause commitments and moving into defense work. Engagement has gone up. Behavior hasn’t. The Index has become the kind of legible instrument that policy shops in Washington and Brussels can point to, which is exactly why the grades matter and exactly why nobody’s clearing a B.
Sources
- https://futureoflife.org/ai-safety-index-summer-2026/
- https://futureoflife.org/wp-content/uploads/2026/07/AI-Safety-Index-Summer-2026-Digital.pdf
- https://time.com/article/2026/07/07/ai-safety-rankings-openai-anthropic-meta/
- https://www.axios.com/2026/07/07/report-ai-safety-pledges
- https://insideaipolicy.com/ai-daily-news/future-life-institute-hands-out-middling-grades-major-ai-developers-safety-efforts