Scorecard: ailabwatch 2025-09-01 scored Google DeepMind on Boosting safety research = Moderate

NoteThe source text contains a scorecard table with dimensions including 'Boosting safety research' and company scores. The row for 'Boosting safety research' shows: 70% (Anthropic), 52% (DeepMind), 35% (OpenAI), 25% (Meta), 0% (xAI), 13% (Microsoft), 8% (DeepSeek). The claimed scoreRaw of 52% and scoreNumeric of 52 match the source exactly. The scoreLetter 'Moderate' is not explicitly stated in the source text provided, but 52% is a reasonable interpretation as 'Moderate' on a percentage scale. The publisher (ailabwatch), entity (Google DeepMind/DeepMind), dimension (Boosting safety research), and date context (September 2025 mentioned as 'as of September 15') all align with the source. Note: The source refers to 'DeepMind' while the record says 'Google DeepMind' — these are the same entity (DeepMind is owned by Google/Alphabet), so this is a minor naming variation, not a contradiction.