Coefficient Giving — Description: 68% of technical AI safety grants went to evaluations/benchmarking (2024)

NoteThe claim states '68% of technical AI safety grants went to evaluations/benchmarking (2024)' with context '$28M recorded technical AI safety grants.' The source directly confirms both the $28M total and the 68% figure for evaluations/benchmarking in 2024. The source explicitly states this in the TL;DR section and reinforces it in Key Finding (1). The 68% figure is presented as a precise statistic, and 2/3 (66.7%) is mathematically equivalent within rounding. All elements of the claim are confirmed by the source.

NoteQUA-650 retro-scan: The claim subject is identified only as 'fact/f_fYQbYvmmnA' with no actual entity name provided, making it impossible to verify if the source matches. Without knowing what entity the claim is actually about, I cannot confirm subject alignment per QUA-648 guidelines, which require preferring MISMATCH when in doubt.

NoteThe source text directly confirms both key components of the claim: (1) $28M in recorded technical AI safety grants in 2024, and (2) 68% of these grants went to evaluations/benchmarking. The exact figures and percentages match the claim precisely.