Training the AI systems: who owns the data? Who bears the risk?
summary
Not all web data is usable for model training: EU law treats a data set as a protectable work so scraping can infringe unless an exception applies (Article 3 for scientific use) or the owner has not opted-out via robots.txt or terms. The EU AI regime forces providers, deployers and general‑purpose developers to document data provenance and build compliance by design; liability distributes across the value chain. US courts use the fair use four-factor test. A recent European case finds nonprofit scientific training lawful.
AI-written summary · about
chapters (6)
- 0:00Introduction to AI data ownership
- 2:33Can all data be used for AI training?
- 7:14Legal risks of unauthorized data use
- 9:00Who is responsible for AI training data?
- 12:12Conclusion on AI training and trust
- 13:14Q&A on web data and jurisdiction
related talks
35mThe European AI Act: a data scientist’s perspective
01h 11mBack to Basics How Data Protection Principles Apply to AI Systems
47mUnderstanding the Legal Context of AI
01h 09mGOVERNANCE AND REGULATION OF AI FROM THE PERSPECTIVE OF AUTONOMY AND PRIVACY
01h 12mAI in the Urban Landscape Navigating Data Governance with Multiple Stakeholders
20mA European Framework for AI
33mAI Regulation in Europe - what's in the pipeline?
38mIs My AI Racist? Or How Not To Break The Law With Machine Learning