The free rider problem in open source AI isn't subtle. Companies train on community contributed data, release nothing back, and call it innovation.
I've watched it happen enough times to stop being surprised by it.
OpenLedger's answer is to make contribution and compensation inseparable at the infrastructure level. If your data enters the training pipeline, the smart contract records it. If the model generates value, the attribution chain traces back to you. No contribution goes unrecorded because the recording is the pipeline.
That's the theory. I spent time looking for the gaps.
The honest gap is enforcement outside the ecosystem. Inside OpenLedger's infrastructure the incentives hold. The moment a trained model leaves that infrastructure, the attribution chain becomes a historical record rather than an active enforcement mechanism.
Records don't stop free riders. They just document them.