Some companies market their generative AI models as open source when they are not open about what data they have used for training them. One such example is Meta’s Llama. It is only open weight. But we are seeing some truly open source generative AI models coming up. Listen to our podcast. And find more here: dataethics.eu/opensource

Shownotes for this episode

• Paper on open washing: dl.acm.org/doi/10.1145/3630106.3659005
• Intellectual Property Cases re GenAI mishcon.com/generative-ai-intellectual-property-cases-and-policy-tracker
• Settlement between Antropic and book publishers: wizcase.com/news/anthropic-wins-ruling-in-copyright-lawsuit
• Data Sheets for Data Sets (2021), Timnit Gebru et al arxiv.org/pdf/1803.09010
• German win on music rights: euronews.com/next/2025/11/11/openai-chatbots-cannot-use-song-lyrics-without-paying-german-court-rules-in-landmark-trial
• A European Open Source Index
• digital-strategy.ec.europa.eu/en/library/europes-open-source-ai-landscape-lever-innovation-and-sovereignty
• Tilde – a european open source foundation language model tilde.ai/tildeopen-llm/
• A new report on the European Open-Source AI Landscape reveals where the EU stands today and how scaling adoption of Open Source AI can boost competitiveness and strengthen digital sovereignty. digital-strategy.ec.europa.eu/en/library/europes-open-source-ai-landscape-lever-innovation-and-sovereignty
• The Model Openness Framework arxiv.org/pdf/2403.13784

Podcast: Play in new window | Download