Do the Future of Life Institute's Summer 2026 AI Safety Index scores match the figures recorded in the vault?
The question
The note claim-fli-summer-2026-index-grades-chinese-labs-below-western records specific grades from the Future of Life Institute's Summer 2026 AI Safety Index: DeepSeek 0.47 (F), Alibaba Cloud 0.87 (D-), Z.ai 0.88 (D-), Anthropic 2.66 (C+), OpenAI 2.28 (C). Do these numbers, letter grades, and lab names match the published index?
Why it matters
Every figure is a load-bearing quantitative claim, and the note's whole argument
(Chinese labs grade far below Western peers despite CnAISDA's institution-building)
depends on the numbers being right. The claim currently rests on a single batch
capture that has not had an independent re-check; the note is held at seedling
and audit_status: capture-verified until this is resolved.
What would answer it
- Read the published index at https://futureoflife.org/ai-safety-index-summer-2026/ directly.
- Confirm the five scores and their letter grades verbatim.
- Confirm the panel framing quotes — "complete passivity" as an existential-safety strategy, and "inadequate safety is a global problem, not a regional one."
- If confirmed verbatim, upgrade the note's
audit_statustoverified-verbatim. If the figures differ, correct the note and flag the discrepancy.
Progress log
- 2026-07-14: Answered by claim-fli-summer-2026-index-grades-chinese-labs-below-western — the 2026-07-09 cross-model (fable) audit independently re-fetched the published FLI index and confirmed all five scores/grades verbatim (DeepSeek 0.47/F, Alibaba Cloud 0.87/D-, Z.ai 0.88/D-, Anthropic 2.66/C+, OpenAI 2.28/C) plus the 'complete passivity' and 'a global problem, not a regional one' framing quotes.