Open-source is NOT the same as open-weight
The distinction between open-source and open-weight is being blurred by mainstream media (including The New York Times) and tech giants. This is a significant mistake for anyone trying to figure out how AI will actually fit into public governance. Open-source software means you have the recipe, including the source code, the training data, and the full pipeline. Open-weight models are just the cake. You can put some icing on it (which people call post-training), but you do not know what is inside.
This distinction matters immensely for public service procurement. If a regional administration buys an open-weight model, they cannot audit the training data for bias or check if the underlying algorithms respect specific privacy regulations. It is a black box with a slightly more accessible front door. You might be able to run it on your own servers, but you cannot truly customize the fundamental behavior of the system because the raw data remains hidden.
But I have to admit that many procurement offices are currently being sold a version of “openness” that is mostly marketing. Companies want the positive press of open software while keeping the proprietary secrets of their training sets. It creates a TRANSPARENCY gap that makes real governance almost impossible.
So we need to stop treating these terms as synonyms in policy documents. For public services to be reliable, we must demand transparency of the training data and the full pipeline, not just the final weights.
#AI #PublicAdministration #Governance #OpenSource #DigitalTransformation
https://garymarcus.substack.com/p/open-source-is-not-the-same-as-open