Scorecard: ailabwatch 2025-09-01 scored Google DeepMind on Scheming risk prevention = Very Weak

NoteThe record claims: publisher=ailabwatch, publishedAt=2025-09-01, entity=Google DeepMind, dimension=Scheming risk prevention, scoreRaw=8%, scoreLetter=Very Weak, scoreNumeric=8. The source confirms all key fields: (1) Publisher is 'AI Lab Watch' (ailabwatch); (2) Date is September 2025 (source states 'Up to date as of September 15' and footer notes 'as of September 2025'); (3) Entity is DeepMind (shown in the scorecard column); (4) Dimension is 'Scheming risk prevention' (explicitly listed as a row); (5) Score is 8% (shown in the DeepMind column for that row); (6) The scoreLetter 'Very Weak' is consistent with an 8% score on a 0-100 scale. The publishedAt date of 2025-09-01 is slightly earlier than the 'as of September 15' update date mentioned in the source, but this is a minor temporal precision difference within the same month and does not contradict the record — the scorecard was maintained throughout September 2025.