Blog

Private data source transparency
2026-10-12 | Private datasources are only required to be explained at a very high level in the public summary of training content. Nonetheless, we observe widespread noncompliance in existing summaries.

Public data source transparency
2026-08-31 | The summaries of public training data are encouraged to have a lot of information on public data sources used. However, we find that in general the information provided is lacking.

Enforcement day is here! 39 public summaries assessed, 20 still missing
2026-08-07 | 2nd August 2026 marks start of enforcement, yet we found 39 Summaries and think 20 are missing

Updates from Providers!
2026-07-10 | Swiss AI Initiative (Apertus) and HuggingFace (SmolLM) have responded to our analysis of their public summaries

We have a blog!
2026-07-01 | We are starting a blog to share regular updates