FOSS, AI & non-code artifacts: What’s new and noteworthy?
summary
Clear rules: separate licenses for code, models, data and media; many model repos show a code license plus an ambiguous “model” license (some even report unknown license for DeepSeek). Practical checks: put a license or notice next to each asset or in a NOTICE file, use CC-BY-SA or CC variants for docs and images, use ODbL for map data, avoid using plain MIT for whole models because MIT license pitfalls only mention “software,” and consider OpenRAIL or revenue‑capped model licenses; use CC0 for SBOMs as SPDX does.
AI-written summary · about
chapters (10)
- 0:00Introduction to FOSS AI licensing
- 2:52Non-code artifacts in FOSS
- 5:50Open source AI definition
- 8:39IP protection for AI outputs
- 13:37Licensing AI models and tools
- 17:55Practical licensing steps
- 21:07AI Act transparency obligations
- 23:14CC licenses for non-code
- 25:21Open RAIL vs OSI compatibility
- 28:35Q&A and closing remarks
related talks
30mOpen Source Licensing: Use Versus Abuse
14mOpen Source Licensing in a Business World
57mWhat You Need to Know About Open Source—Trust Me, I'm a Lawyer
01h 38mOpen Source Licensing: Types, Strategies and Compliance
42mWhy and How We Updated Our License Policy for Use of Open Sou...
45mOpen Source distributions in a cloud-native world: from a technical to a legal point of view
42mOpen Source Licenses: Between Parish Wars and Radicalism
14mOpen-source it right (and avoid calling your IP lawyer)!
