The big debate going on in the artificial intelligence (AI) is around Open Weights, and whether AI suppliers should embrace the concept, or make their models more restrictive.
But what are Open Weights? Is it equivalent to Open Source AI? And why is the debate important?
According to the Open Source Initiative, Open Weights indicate incremental progress in AI transparency. By sharing the final parameters of a trained model, developers offer some insight into how a neural network operates.
However, the OSI cautions that these weights reveal only a fraction of the information required for full accountability.
“While Open Weights represent a milestone in opening up AI systems, they still stop short of delivering the level of transparency many researchers and regulators deem essential.”
The organisation explains that Open Weights refer to the final weights and biases of a trained neural network. These values, once locked in, determine how the model interprets input data and generates outputs.
When AI developers share these parameters under an OSI Approved License, they empower others to fine-tune, adapt, or deploy the model for their own projects.
However, Open Weights differ significantly from Open Source AI because it does not include training code, the training dataset or comprehensive data transparency.
By withholding these critical elements, developers only provide a glimpse into the final state of the model, making it difficult for others to replicate, audit, or deeply understand the training process, according to the Open Source Initiative.
While Open Weights is more transparent than proprietary, the body points to other elements of Open Source AI that are still missing, including lack of reproducibility; data opacity; regulatory hurdles; and limited community collaboration.
Without the entire pipeline – training scripts, dataset composition and intermediate checkpoints – openly available, Open Weights limits the global community from making meaningful contributions to superficial fine-tuning rather than in-depth improvements.
The body adds that, while Open Weights is a big improvement on a proprietary model, it is not as comprehensive as Open Source AI, which is made available under terms that grant users freedom to use, study, modify and share.
There are also ethical consideration, according to the Open Source Initiative: Open source AI allows reviewers to spot and address potential biases early; while Open Weights alone can’t provide enough context to guarantee ethical performance.
This is relevant in terms of regulatory compliance, innovation and collaboration, and trust and public perception.
The organisation concludes that Open weights doesn’t go far enough.
“Many see Open Weights as a compromise—a lesser evil than completely proprietary AI. By at least making the final parameters accessible, developers provide some degree of insight into the model’s decision logic. This can be enough for certain low-stakes applications where minimal accountability suffices.
“However, for industries like healthcare, autonomous vehicles, or financial underwriting – where AI decisions carry significant consequences – the partial transparency of Open Weights is insufficient. Full accountability demands understanding not just the final model, but also how it was built, the data it relied on, and the points at which it might have diverged from ethical best practices..
The full statement from the Open Source Initiative can be seen here.