Artificial Intelligence

Top AI Firms Earn Failing Grades on FLI Safety Index

Is AI safety grading a necessary accountability tool or an oversimplification of a genuinely complex field?

Published JUL 8

Top AI Firms Earn Failing Grades on FLI Safety Index

Story

Above: A smartphone screen displays a folder containing AI applications on Feb. 6. Image credit: Samuel Boivin/NurPhoto/Getty Images

The Facts

The Future of Life Institute on Tuesday released its 2026 first-half AI Safety Index, evaluating nine major AI companies — Anthropic, OpenAI, Google DeepMind, Meta, Z.ai, Alibaba Cloud, xAI, DeepSeek and Mistral — across six categories including risk assessment, existential safety and governance.
BigGo Finance
The Straits Times
Future of Life Institute on X
Anthropic ranked first overall with a grade of C+, while OpenAI and Google DeepMind each received a C and Meta scored a D+. xAI, DeepSeek and Mistral all received failing grades, and not a single company earned an A or B in any category.
Seoul Economic Daily
Inside AI Policy
The index found that Anthropic, OpenAI, Google DeepMind and Meta had each weakened or abandoned prior commitments to halt development if certain risk thresholds were reached. Anthropic dropped a pledge in February not to train AI systems without first demonstrating its safety measures were sufficient.
WBUR
Time
FOX News
The FLI report identified the expansion of military AI applications as a serious concern, noting that major companies had reversed earlier bans on military use. Anthropic was specifically criticized for what the report called "questionable military engagement," citing media reports that its technology was used in U.S. military operations in Venezuela and Iran.
BigGo Finance
Axios
Existential safety was the weakest category across all nine companies, with the lowest-scoring companies — xAI, DeepSeek and Mistral — hailing from the U.S., China and Europe, respectively.
Seoul Economic Daily
Time
France 24
Mistral, included in the index for the first time, finished last and disputed the result, saying the evaluation framework was not suited to its open-weight model approach. The FLI noted that Mistral, along with Alibaba, xAI and DeepSeek, did not respond to its survey.
Future of Life Institute
The Straits Times

Sources Split

LEFT

CENTER

RIGHT

ESTAB
CRIT

CENTER

PRO
ESTAB

The Spin

Establishment-critical narrative

Nine of the biggest AI labs just got graded on safety and governance — and not one passed. Even more alarmingly, four abandoned pledges they made in existential risk. This is what happens when competitive pressure meets self-policing. Frontier AI buyers can't just take model cards at face value anymore; they need hard answers on risk thresholds, access to external testing and incident reporting before deploying these systems.

Future of Life Institute on X

Let's Data Science

Pro-establishment narrative

AI safety is a complex, evolving field with no universal standard — terms like alignment, red-teaming and risk thresholds mean different things to different developers. Grading frontier labs on governance without accounting for the technical nuance behind safeguards, evaluations and deployment contexts oversimplifies a deeply layered challenge.