case against osaid

The Case Against OSI's Open Source AI Definition

⚠️ Sample. An AI-generated demo of LLM Wiki Newsroom — the "open source AI" topic is just the example corpus.

Summary

David Cassel rounds up the growing criticism of the Open Source Initiative's Open Source AI Definition, centered on its decision not to require open training data. Critics — including OSI co-founder Bruce Perens, the Software Freedom Conservancy's Bradley Kuhn, the Free Software Foundation, RedMonk's Stephen O'Grady, and OpenUK's Amanda Brock — argue the definition erodes the meaning of open source, with some warning it threatens the future of the term itself.

Key Claims

• [fact] OpenSourceInitiative — approved the OSAID through its 10-person board rather than a full membership vote
• [analysis] FreeSoftwareFoundation — holds that a machine-learning application is not free unless its TrainingData and processing scripts respect the four freedoms
• [analysis] OpenSourceInitiative — defends omitting full training data so that OpenSourceAI can exist in fields like medical AI where data cannot be legally shared
• [forecast] FreeSoftwareFoundation — Bradley Kuhn announced a campaign to run for the OSI board on a platform to repeal the OSAID

Key Quotes

"I don't think the Open Source AI Definition was a good idea at the start, and it hasn't turned out well. The plain old Open Source Definition that we've had for 26 years can be applied to AI." — OpenSourceInitiative

Connections

• contradicts: OpenSourceAI — argues the OSAID dilutes the open-source brand and is "less than Open Source"
• references: TrainingData — the omission at the heart of the criticism
• references: OpenSourceInitiative — the body whose definition is contested
• references: FreeSoftwareFoundation — developing a stricter, open-data alternative
• references: OpenWashing — critics fear the definition enables open-washed models
• references: ModelLicensing — the dispute is over what licensing terms count as open