Seven independent experts graded every frontier AI lab on safety. The best grade was a C+. Three labs got an F. The best existential-safety grade was a D+, and nobody reached a C. And the labs that used to promise to stop at red lines? They rewrote the red lines.

SAFETY FUTURE OF LIFE INSTITUTE (OFFICIAL AI SAFETY INDEX, SUMMER 2026) · CLAIMED 2026-07-1904/7

Seven independent experts graded every frontier AI lab on safety. The best grade was a C+. Three labs got an F. The best existential-safety grade was a D+, and nobody reached a C. And the labs that used to promise to stop at red lines? They rewrote the red lines.

The Future of Life Institute's Summer 2026 AI Safety Index scores nine frontier labs across 37 indicators. Anthropic leads at C+ (2.66 out of 4.0). xAI scored 0.65, DeepSeek 0.47, Mistral 0.33. Four companies did not even respond to the survey. The panel found that labs which once pledged to pause at danger thresholds have replaced those commitments with competitor-contingent conditions: I will stop if they stop, which means nobody stops.

01THE CLAIM

"Every frontier AI lab scores C+ or below on the FLI AI Safety Index; no company passes on existential safety." [SOURCE ↗]

VERIFIED5 SOURCES · LIVE 2026-08-25

FUTURE OF LIFE INSTITUTE TRACK RECORD1 CLAIM · 0/100 BS RATE →

C+Anthropic's grade, the HIGHEST of any frontier lab (2.66/4.0)

FxAI (0.65), DeepSeek (0.47), and Mistral (0.33) all fail

0companies reaching even a C in existential safety (best was D+)

4of 9 companies that did not even respond to FLI's survey

Seven independent experts graded every frontier AI lab on safety. The best grade was a C+. Three labs got an F. The best existential-safety grade was a D+, and nobody reached a C. And the labs that used to promise to stop at red lines? They rewrote the red lines.

02THE CHECK

THE CLAIM. Every frontier AI lab scores C+ or below on the FLI AI Safety Index. No company passes on existential safety.

THE CHECK. seven experts from UC Berkeley, Oxford, U Montreal, and others evaluated nine labs on 37 indicators across six domains through June 2026. The panel found Anthropic, OpenAI, Google DeepMind, and Meta weakened or voided pledges to pause if redlines are approached, replacing them with competitor-contingent conditions. Labs that once banned military applications now seek defense contracts. David Krueger called the lack of progress scandalous. Four of nine companies did not respond to the survey.

THE PATTERN. safety pledges conditional on competitors matching them are pledges that expire on contact with competition. The industry grades at C+ at its best and F at its worst, while seeking trillions in public market valuation.

03SAY THIS IN THE MEETING · 📸 SCREENSHOT IT

"The best safety grade any frontier AI lab received from independent experts is a C+. Three labs got F. The best existential-safety grade was Anthropic's D+, and nobody reached a C. The labs that pledged to stop at danger thresholds have rewritten those pledges to say they will only stop if competitors stop first."

The Future of Life Institute has been grading frontier AI labs twice a year since 2024. The Summer 2026 edition evaluated nine companies across six domains: risk assessment, current harms, safety frameworks, existential safety, governance and accountability, and information sharing.

🔒 THE FULL AUTOPSY · FREE WITH AN ACCOUNT

You just read the free check. Sign in free, a code by email, no passwords, and the rest unlocks: the evidence trail, the steelman and the rebuttal, all 5 sources with quotes and screenshots, and our on-record call.